Download QUANTITATIVE TEXTURAL MEASUREMENTS IN IGNEOUS AND

Document related concepts
no text concepts found
Transcript
This page intentionally left blank
QUANTITATIVE TEXTURAL MEASUREMENTS
IN IGNEOUS AND METAMORPHIC PETROLOGY
Processes involved in the development of igneous and metamorphic rocks
involve some combination of crystal growth, solution, movement and deformation, which is expressed as changes in texture (microstructure). Recent
advances in the quantification of aspects of crystalline rock textures, such
as crystal size, shape, orientation and position, have opened new avenues
of research that extend and complement the more dominant chemical and
isotopic studies.
This book discusses the aspects of petrological theory necessary to understand the development of crystalline rock texture. It develops the methodological basis of quantitative textural measurements and shows how much can be
achieved with limited resources. Typical applications to petrological problems
are discussed for each type of measurement. The book has an associated
webpage with up-to-date information on textural analysis software, both
commercial and free. This book will be of great interest to all researchers
and graduate students in petrology.
M I C H A E L H I G G I N S is Professeur des Sciences de la Terre at the Université
du Québec à Chicoutimi, Canada. He is a member of the Geological
Association of Canada.
QUANTITATIVE TEXTURAL
MEASUREMENTS IN IGNEOUS AND
METAMORPHIC PETROLOGY
MICHAEL DENIS HIGGINS
Sciences de la Terre,
Universite´ du Québec à Chicoutimi,
Canada.
  
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press
The Edinburgh Building, Cambridge  , UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521847827
© M. D. Higgins 2006
This publication is in copyright. Subject to statutory exception and to the provision of
relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
First published in print format 2006
-
-
---- eBook (EBL)
--- eBook (EBL)
-
-
---- hardback
--- hardback
Cambridge University Press has no responsibility for the persistence or accuracy of s
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
To Ester and Zoe
Contents
page ix
Acknowledgements
1 Introduction
1.1 Petrological methods
1.2 Qualitative versus quantitative data
1.3 What do I mean by texture?
1.4 Information density and data sources
1.5 Structure of this book
1.6 Software applications for quantitative textural studies
2 General analytical methods
2.1 Introduction
2.2 Complete three-dimensional analytical methods
2.3 Extraction of grain parameters from data volumes
2.4 Destructive partial analytical methods
2.5 Surface and section analytical methods
2.6 Extraction of textural parameters from images
2.7 Calculation of three-dimensional data from
two-dimensional observations
2.8 Verification of theoretical parameter distributions
2.9 Summary
1
1
2
3
4
5
5
7
7
10
13
14
16
26
33
35
38
39
39
42
74
102
3 Grain and crystal sizes
3.1 Introduction
3.2 Brief review of theory
3.3 Analytical methods
3.4 Typical applications
vii
viii
Contents
4 Grain shape
4.1 Introduction
4.2 Brief review of theory
4.3 Methodology
4.4 Typical applications
135
135
136
143
154
5 Grain orientations: rock fabric
5.1 Introduction
5.2 Brief review of theory
5.3 Introduction to fabric methodology
5.4 Determination of shape preferred orientations
5.5 Determination of lattice preferred orientations
5.6 3-D bulk fabric methods: combined SPO and LPO
5.7 Extraction of grain orientation data and parameters
5.8 Typical applications
165
165
166
170
171
174
177
181
184
6 Grain spatial distributions and relations
6.1 Introduction
6.2 Brief review of theory
6.3 Methodology
6.4 Typical applications
199
199
200
204
212
7 Textures of fluid-filled pores
7.1 Introduction
7.2 Brief review of theory
7.3 Methodology
7.4 Parameter values and display
7.5 Typical applications
216
216
217
222
226
228
8 Appendix: Computer programs for use in quantitative
textural analysis (freeware, shareware and commercial)
8.1 Abbreviations
8.2 Chapter 2: General analytical methods
8.3 Chapter 3: Grain and crystal sizes
8.4 Chapter 4: Grain shape
8.5 Chapter 5: Grain orientations (rock fabric)
8.6 Chapter 6: Grain spatial distributions and relations
232
232
232
235
235
236
236
References
Index
237
260
Acknowledgements
The inspiration for this book came from a short course that I gave in September
2001 in Ankara, Turkey. I would like to thank Durmus Boztug, Cumhuriyet
University, Bahadir Sahin, MTA Anakara, the Turkish Chamber of Geological
Engineers, TUBITAK and MTA, Anakara for making that short course possible. This book was started while I was on sabbatical at Victoria University,
Victoria, Canada and I would like to thank the department, and particularly
Dante Canil for hosting me. I would like to thank my readers for their comments: Marian Holness, Cambridge University, Ned Chown, Paul Bédard and
Pierre Cousineau, Université du Québec à Chicoutimi, Dougal Jerram, Durham
University, Lawrence Coogan, University of Victoria, Keith Benn, University
of Ottawa, Alison Rust, University of British Columbia, and Judit Ozoray.
J. M. Derochette drew my attention to the Benson plate. Data were supplied by
Ilya Bindeman, Margaret Mangan and Ronald Resmini. Patrick McLaren of
Geosea Consulting, Victoria, showed me how sediment sizes are measured.
I would also like to acknowledge Bruce Marsh, Johns Hopkins University,
and Kathy Cashman, University of Oregon, without whom CSD research
would never have moved along so fast. The Natural Science and Engineering
Research Council of Canada and the Fondation de l’UQAC have supported my
research and have enabled many of the studies presented here. Last I would like
to thank my colleagues at UQAC, my students and my family.
ix
1
Introduction
I do not know what I may appear to the world, but to myself I seem
to have been only like a boy playing on the sea-shore, and diverting
myself in now and then finding a smoother pebble or a prettier shell
than ordinary, whilst the great ocean of truth lay all undiscovered
before me.
Sir Isaac Newton
Newton’s ‘Ocean of Truth’ seems to me more like a landscape: the plains are
densely populated with information and most researchers work there. It is not
easy to get a perspective on such a mass of information without climbing the
surrounding hills. Valleys in the mountains may be difficult to find, and
sparsely populated, but some lead to new basins of information ready to be
explored. Other valleys may be so deep that we can glimpse what they contain,
but cannot explore them closely, or even at all. This book is a guide to a
country set in that landscape. Like real countries, its name varies according to
who you ask, and the borders do not always follow geographic features. And
like most travel writers, my description of the landscape is coloured by where
I come from and what I have done.
1.1 Petrological methods
In petrology1 we generally examine the results of natural experiments and have
to tackle the inverse problem of what happened to what starting material to
produce the rock that we observe. This problem is complex and does not
usually have a unique solution: we must decide what the most likely starting
material was and what the dominant processes were. The general approach
is to choose a number of parameters and quantify them in the final rock.
A starting material is proposed and various processes applied to this material.
The results of this modelling are then compared to the rock in question.
1
2
Introduction
All rocks comprise an assemblage of mineral grains of different sizes,
orientations and shapes, which may be observed in many different ways.
Early petrological studies of rocks used observations in the field and with a
polarising microscope. The value of chemical analyses was appreciated, but
the effort involved to analyse a small number of samples was considerable. The
development of instrumental methods of chemical and isotopic analysis changed this and now such methods have become very important. In fact, many
geologists view igneous and metamorphic petrology as simply applied mineral,
chemical and isotopic analysis. This is a rather narrow view and in this book
I hope to show that quantitative analysis of textures can also help unravel the
evolution of materials. In a way, textural methods are more direct than
geochemical or isotopic methods as the most important processes involve
changes in the size, shape, position and orientation of the crystals. However,
it is better to view all these methods as complementary.
This book is aimed at researchers in igneous and metamorphic petrology at
all levels. Some workers may not be aware of the potential of textural studies
to solve problems that they have previously attacked only with chemical or
isotopic methods. Others may have started textural studies, but are unaware of
what has been done in other subfields of petrology.
An ideal petrological study would combine relevant theory, appropriate
analytical methods and application to real rocks. However, petrologists are
human and in many cases one aspect is developed to the detriment of the other:
the siren calls of theory and methods seem to be heard more often than
application to rocks. I hope that readers will be able to combine both highquality measurements with well-grounded theory to advance our understanding of the origin of rocks.
1.2 Qualitative versus quantitative data
Although the patterns and structures of rocks have been observed and recorded
since early times, it was only in the nineteenth century that the development of
the polarising microscope enabled a vast increase in the detail and number of
petrological studies. At that time such observations were mostly qualitative –
such as average grain-sizes, grain relationships, grain boundary shapes and
orientation fabrics. While qualitative data have their uses, they cannot constrain
physical models of processes in the same way that quantitative data can (for a
history of quantification in geology see the review of Merriam, 2004). Models
can be developed that make quantitative predications which can only be verified
if accurate quantitative observations of rocks are available. All too often
authors have developed elaborate mathematical models of geological processes,
1.3 What do I mean by texture?
3
only to remark at the end of the article that their results qualitatively resemble
actual rocks. What are clearly needed are quantitative measurements of rock
textures that can be compared to theoretical predictions.
The division between qualitative and quantitative data is somewhat artificial and relates to aspects of data quality. In quantitative studies data quality
refers largely to the accuracy of the measurements. If data are not accurate
then they may be no more useful than qualitative data and indeed may
be detrimental. In good geochemical studies accuracy is proved by analysis
of geochemical reference materials. Although there are a few such
materials available for textural analysis they are of limited application and
are rarely used.
1.3 What do I mean by texture?
The texture of a rock is considered here to be the geometric arrangement of
grains, crystals, bubbles and glass in a rock. I use the term rock here in its most
general sense: for both lithified and loose materials. Materials scientists and
geophysicists also use the term polycrystal to describe the same sorts of
materials. Although the internal structure of grains and crystals is very important, and may help our interpretations, it is not developed here. The term
texture as used here includes orientation of crystals, which is sometimes
separately referred to as fabric. In other fields the terms fabric and texture
are used in the opposite way. In materials science the term microstructure
covers the same field (Brandon & Kaplan, 1999) and this is favoured by some
geologists (Vernon, 2004). However, the name of the subject is not really
important, as compared to its content.
A number of textural components can be, and are commonly quantified:
*
*
*
*
*
Size of grains, crystals and bubbles.
Shape and boundary convolutions of grains and crystals.
Orientation of grain and crystal shapes and crystal lattices.
Position and connectedness of crystals.
The relationships between different phases.
Other textural components, such as enclosure relationships, are important but
are not usually quantified. They can be useful as earlier textures may be
preserved within oikocrysts or other structures (e.g. Higgins, 1998).
The precision and accuracy of textural parameters determined from rocks is
very variable, but does not correlate with the relevance of these parameters to
current problems in petrology. However, if a parameter can be readily and
accurately measured, then it should be reported or made available in an
4
Introduction
electronic archive, even if its relevance is not apparent when the study is done:
we have little idea what will be needed for future studies.
Quantitative textural analysis of rocks must start with measurement of
textural parameters in rocks. Modelling of the evolution of these parameters
is generally done by taking each parameter individually. For example crystal
size is generally modelled independently from other textural parameters (e.g.
the classic study of Cashman & Marsh, 1988). This probably reflects on the
youthful nature of the field, as compared to geochemical or isotopic studies.
However, there have been some attempts to model the whole 3-D texture of a
rock mathematically.
1.4 Information density and data sources
One of the problems inherent in a book such as this is the huge variation in the
number of studies in different branches of the subject. I call this the information density. For example, there are hundreds, or maybe thousands, of studies
of the orientation of crystals in metamorphic rocks, yet probably less than ten
in igneous rocks. This means that in some fields I have just summarised the
field, referred to a specialised text or discussed a few key studies, whereas in
others I mention all work in the subject. A subjective assessment of information density is shown in Table 1.1.
One of the most important roles of this book is to broaden the application of
methods across the subject, and hence to even out these variations in information density.
Scientific data are circulated in many forms: some are rigorously refereed and
easily accessible, whereas others do not meet these criteria. In this work I have
not referenced or used short conference abstracts as they are not really refereed
and rarely contain enough information to be useful. Often they seem to be just
marking out territory. Many M.Sc. and Ph.D. theses do contain useful data
and methods, but they are not refereed to the same standards as most journal
Table 1.1 Estimates of information density in quantitative textural studies:
O ¼ little work, OOO ¼ many studies.
Field of study
Igneous petrology
Metamorphic petrology
Crystal and pore size
Crystal shape
Crystal orientation
Crystal position
OO
O
O
O
O
O
OOO
O
1.6 Software applications
5
papers. In addition, theses are not always easy to come by and their existence
and content tends to be passed on by word-of-mouth. Hence, I have generally
omitted theses: high-quality studies will generally be published anyway by the
student or the supervisor.
1.5 Structure of this book
Although this book covers a wide range of quantitative textural approaches to
petrological problems many share similar analytical methods. Hence, I start in
Chapter 2 with such generally applicable methods. I will then divide the subject
on the basis of the parameters measured: size, shape, orientation and position.
In Chapters 3 to 6 I will briefly review enough theory so that the nature of the
problems can be understood. I then continue with a survey of the analytical
methods, where they differ from the more generally applicable methods discussed in Chapter 2. The data produced by some analytical methods require
significant treatment to extract meaningful textural parameters. I will present
some applications that are typical of this subject, innovative, or which show
the way that the subject may develop in the future. If I make too many
references to my work and that of my students, then it is because I am familiar
with the data, and can recalculate them to compare them with the data of
others. Finally, I finish with Chapter 7 on all aspects of the textures of pores.
1.6 Software applications for quantitative textural studies
The analytical methods described in this book use a variety of different software applications. Many researchers are unwilling or unable to invest the time
and money needed to acquire new software and the skills to use it. This is
particularly evident if the software is not free-standing, that is, it must be used
with other applications. These factors result in a tremendous range in the
application of new methods that is not entirely dependent on their overall
scientific usefulness. A caveat for researchers is that if you want your method
to be widely employed, then you must make it easy and cheap to use. Software
types and mathematical procedures are presented below in decreasing order of
likelihood of use.
1. General-purpose image processing and spreadsheet programs (e.g. Adobe Photoshop
and Microsoft Excel). These are available for a number of operating systems
(Windows, Mac, Linux, etc.). A lot of data reduction can be done with such
programs, but they are commonly oriented towards business applications and
hence may involve rather complex manipulations to do simple calculations.
6
2.
3.
4.
5.
6.
Introduction
Sometimes it is easier to augment these programs by writing a small program to do
one step in the processing or to convert data formats.
Special programs: freeware, shareware and commercial. Some programs are readily
available (e.g. CSDCorrections, NIHImage) while others can be very expensive.
Most freeware is not available in all operating systems. However, Java programs
(e.g. ImageJ) can be used on most operating systems. There are a number of large and
comprehensive statistical packages that can be used to manipulate data (SAS, SPSS).
Macros (procedures) for high-level languages, such as Mathematica, Maple, Matlab
(Middleton, 2000), ImageJ and IDL. Such languages are very powerful but are not
so readily available. They have a steep learning curve for beginners. Many academic researchers may not realise that they already have access to these packages,
sometimes via engineering departments. Many compiled IDL programs can be run,
but not modified, with the aid of a free program called a ‘Virtual Machine’.
Uncompiled program code in common languages such as FORTRAN, C, Pascal or
Basic. Such code must be compiled for the operating systems that will be used.
Many compilers are commercial and can be expensive. Minor code modifications
are commonly needed and necessitate some knowledge of programming or a helpful assistant.
Mathematical equations and algorithms. Some equations are easily applied but
others necessitate that software must be written before the method can be applied.
‘Home-made’ software. Software development with ‘visual’ compilers like Visual
Basic, Visual C or Delphi (¼ visual Pascal) is relatively easy for simple applications,
but difficult to perfect. If you have no programming experience you should expect
to spend a week before you can make a simple program. Visual Basic is slow and
should not be used for calculation intensive applications. Visual C or Delphi are
not much more difficult to learn and produce a faster program.
All methods should ideally be tested with standard materials or synthetic data
derived using independent methods, but this is not always available. Software
‘bugs’ and operating system incompatibilities are inevitable – producers of
scientific software cannot usually test their programs on all systems and generally
welcome constructive comments. They are generally geologists and not programmers. They do not like to hear ‘your program never worked so I abandoned it’.
Software mentioned in this book is described in the Appendix and also on a
website, which will be updated regularly (http://geologie.uqac.ca/%7Emhiggins/
CSD.html).
Notes
1. In this book I will address igneous and metamorphic petrology. Quantification of textures is
an important part of sedimentary petrology, particularly grain-size analysis of unconsolidated sediments, but the subject is too vast and distant from the petrology of crystalline rocks
to be bridged by one author in one book.
2
General analytical methods
2.1 Introduction
2.1.1 Observations in three and two dimensions
Many different analytical methods have been applied to quantifying the
textures of rocks, but all start with observations of rocks either in three or
two dimensions. In this chapter I will start with general analytical methods
that apply to the determination of many different textural parameters.
Methods specific to a particular textural parameter will be discussed later in
the relevant chapters. It should be remembered that different methods can be
combined within a study. This can validate data acquired by innovative
methods and also extend the range of grain sizes that can be quantified
(Figure 2.1). Some methods that I will describe are used widely (e.g. transmitted light microscopy) whereas others are discussed here because I think that
they have potential for future textural studies (e.g. Nomarski imaging).
Researchers do not always choose their analytical method in the most logical
way: here I want to survey the field to show what others have done. However,
researchers, especially students, should never forget that a lot can be done with
little equipment – my first crystal size distribution study used a microscope,
camera, ruler and protractor (Higgins, 1991).
The texture of many different objects or parts of objects in a rock can be
described: it is important to specify at the start what you want to examine. For
instance, if a crystal is fractured then the size of the fragments or the size of the
original crystal can be measured. If grains are altered at their rims then the
original size of the grain or the present size of the grain can be measured.
The texture of rocks is a three-dimensional (3-D) property, hence it is most
directly studied by analytical methods that view blocks of rock. However, such
methods are not always applicable: samples, and grains, may be too large or
small for 3-D methods; the equipment may not be available or too expensive
7
8
General analytical methods
m
m
0.
00
01
m
m
0.
00
1
m
m
0.
01
m
m
0.
1
1
m
m
m
m
10
10
10
0
00
m
m
m
m
typical grain or crystal sizes
Electronic methods – SEM; EMP;
TEM; cathodoluminescence
Thin sections –
reflected light
Thin sections –
transmitted light
Blocks – X-ray tomography/
serial sectioning
Slabs – direct
measurement or
image analysis
Outcrop surfaces –
direct measurement
m
1
00
0.
0.
01
m
m
m
m
1
0.
1
m
m
m
m
m
10
m
10
0
m
m
m
00
10
10
00
0
m
m
Sample solution/disintegration
linear dimensions of sample
Figure 2.1 Quantitative textural methods that can be used at different scales
of interest. Electronic methods include scanning electron microscopy (SEM),
electron microprobe (EMP), transmission electron microscopy (TEM) and
cathodoluminescence. The scale is the linear dimension of the sample – the
edge of the surface or block. Maximum grain or crystal sizes that can be
measured successfully are typically about one tenth of these values. Not all
methods can give all textural parameters: for example, sample solution
cannot give information on crystal position or orientation. Even within the
scale range of a single method it may be necessary to gather the information at
two different magnifications.
for the study; and many 3-D methods do not distinguish clearly touching
grains of the same phase. Hence, it is commonly necessary to study sections
through a rock, and extrapolate these data to three-dimensional textural parameters. The branch of mathematics that deals with this problem is known as
stereology (Underwood, 1970). Detailed stereological methods will be discussed in the relevant chapters. However, here I will mention a number of
‘global parameters’, which describe uniquely various geometrical properties of
2.1 Introduction
9
a solid and can be determined from examination of a section. Finally, I will
briefly describe some of the more general methods used to summarise the
distribution of textural parameters, whether determined by 2-D or 3-D analytical methods.
2.1.2 Grain size limits
The size limits and resolution of different analytical methods must be respected
(Figure 2.1). The resolution of many analytical methods is clear – for twodimensional images it is commonly the size of the pixels (picture elements –
dots or groups of different coloured dots on the screen). However, other
factors may reduce the resolution, such as the lenses or physical limits. The
minimum size of grain or crystal that can be measured accurately must be
established from this value. For size measurements a typical limit might be ten
times the resolution: for example, each measured crystal must have a length
of at least 10 pixels. If a method uses discrete intervals, then there must also
be sufficient grains or crystals in an interval such that the value is statistically
meaningful. For example, if a size interval contains only one crystal then
that value has very little meaning. These factors generally limit the maximum
size of grains or crystals that can be quantified to about one tenth of the size of
the image.
2.1.3 Edge effects
Almost all analytical methods, both 3-D and 2-D, must deal with the problem
of edge effects of samples. Boundaries of a sample commonly intersect some
crystals, which must be dealt with in a consistent manner. If the sample
contains many crystals then edge effects can be ignored. For smaller samples
two simple solutions are possible. The area or volume of measurement need
not have a simple shape and can be a complex envelope that only includes
completely visible crystals (Figure 2.2a). The area or volume of measurement
can be a simple rectangle or parallelepiped that is drawn inside the actual limits
of the sample so that partially visible crystals are excluded (Figure 2.2b). Then
only those crystals that touch two sides of the square, or three edges of the
parallelepiped, are included.
2.1.4 Textural development sequences
When a rock is sampled all that can be generally observed is the final product –
the path taken to achieve that product is not always clear, but may sometimes
10
General analytical methods
(a)
(b)
Figure 2.2 Two solutions to the problem of edge effects in textural measurements, shown here for a section. (a) Crystals that are not completely visible in
the field of view (open outlines) are excluded and an irregular envelope
passing midway between the crystal edges (dotted line) encloses the crystals
to be measured (grey outlines). The area of the envelope is the area measured.
(b) A rectangle is drawn around all crystals that are completely visible.
Crystals are counted that fall within the rectangle or touch two of the sides.
Those that touch the other two sides (dashed) are excluded as well as those
completely outside the rectangle.
be revealed by carefully examining other, less developed samples, or by locating early textures frozen in by other processes. Such sequences of textural
development can help enormously in understanding petrogenetic processes.
For instance a series of lavas or samples taken from a lava lake may show how
magmas crystallise (e.g. Cashman & Marsh, 1988). First-formed plutonic
textures may be seen in oikocrysts and early metamorphic textures may be
preserved in porphyroblasts (e.g. Higgins, 1998). Hence, the methods
described below should be applied not only to the average rock, but also to
special sectors of a sample that can show evidence of earlier textures.
2.2 Complete three-dimensional analytical methods
Three-dimensional analytical methods that conserve the size, shape, orientation and position of the crystals will be discussed first. Some of these methods
conserve the sample (e.g. X-ray tomography) whereas others are destructive
(e.g. serial sectioning).
2.2.1 Serial sectioning
The complete texture of a sample can be established from serial sectioning
(Bryon et al., 1995). Here, a surface or thin section is cut and recorded as a
photograph or digital image. The surface is then ground away and a new
2.2 Complete three-dimensional analytical methods
11
surface or thin section made, parallel to the original section, and the process
repeated. Clearly, the sample is either destroyed or reduced to a series of
thin sections. The resolution of the method is limited by the spacing of the
sections and the resolution of each image, which should ideally be equal
(Marschallinger, 1998b, Marschallinger, 1998a). The images can be processed
as for surface methods to separate out the different minerals (see Sections 2.6.2
and 2.6.3). The processed images can then be combined into a data volume (3-D
image) to establish the complete shape of each crystal (Marschallinger, 2001).
Serial sectioning can give excellent results, especially for small numbers of
irregularly shaped objects, but it is very time consuming and its resolution is
limited by the spacing of the sections. For example, 500 successive images must
be obtained and combined to match an image with a resolution of 500 500
pixels. This is rarely done and the vertical resolution is generally much less
than the resolution in the plane of the images. If the sample is ground away
(‘lapped’) to make the separate images then a resolution as small as 40 mm has
been achieved (Marschallinger, 1998a). If thin sections are used then a much
larger spacing is needed, typically several mm. Of course, the vertical resolution must be balanced by the need and ability to distinguish individual crystals.
Some textural parameters do not need crystals to be separated (e.g. intercept
orientation method) and in some rocks crystals can be readily isolated in plane
surfaces without optical orientation. However, if crystals must be separated
then it is generally easier to do in a thin section than on a flat surface.
Serial sectioning has been used more extensively in biological and palaeontological studies where the interest is in small numbers of very irregular
objects – whereas petrology is more unusually concerned with large numbers
of similarly shaped objects, like crystals.
2.2.2 Optical scanning and confocal microscopy
Optical scanning and confocal microscopy are special techniques that can be
used for the measurement of small proportions of grains in transparent materials. They give a result similar to serial sectioning, but without destroying the
sample. In optical scanning the section is examined at high magnification with
a large aperture. In this situation the depth of field is small and a narrow range
of depths in the section are focused. A photograph is taken and the sample–
objective distance increased. The process is repeated to build up a complete
3-D reconstruction of the section. The matrix of the crystals must be sufficiently
transparent and the crystal number density must be sufficiently low that the
whole crystal can be observed; hence it can only be applied in special circumstances, for instance microlites in a glassy volcanic rock (Castro et al., 2003).
12
General analytical methods
A confocal microscope is designed specifically for these types of application
(e.g. Petford et al., 2001, Bozhilov et al., 2003). Instead of shining a light on the
whole section, only a part of the sample is illuminated with a laser beam that is
scanned across the sample. The image is then reconstructed sequentially, as in
a scanning electron microscope (see below). This method has the advantage
that scattering of light from adjacent crystals and matrix into the volume of
interest is much reduced.
It is not always necessary to reconstruct the whole 3-D structure: the length
and other shape parameters can be determined from the vertical and horizontal position of the ends of the crystal. In some cases the method has been
simplified further by choosing crystals that are nearly parallel to the plane of
the section (Castro et al., 2003).
2.2.3 X-ray tomography
Tomography (CAT – computed axial tomography) is the reconstruction of a
section from many separate projections around an object. A series of closely
spaced slices are assembled into a 3-D image. It is commonly applied using
X-rays (Ketcham & Carlson, 2001, Mees et al., 2003), but can also be used with
any radiation that can penetrate the material, such as gamma rays or even light
(Figure 2.3).
X-ray
detectors
X-ray
source
Object
Figure 2.3 X-ray tomography. In most systems used for geological research
the sample is rotated as it is bathed in a flat beam of X-rays. The X-rays that
pass through the sample are detected by a curved bank of detectors. A slice of
the internal structure of the sample can be reconstructed from the quantity of
X-rays received by the detectors. The sample is then moved vertically and
another slice analysed.
2.3 Extraction of grain parameters
13
Minerals are distinguished in X-ray tomography on the basis of their linear
attenuation coefficient, m. This depends directly on the density of the mineral,
the effective atomic number of the mineral, and the energy of the incoming
X-ray beam. Common minerals have values of m that vary by a factor of three
for 100 keV X-rays, which is much greater than the precision of measurement,
typically 0.1%. However, there is much overlap in the values of m, hence
minerals cannot always be distinguished using this parameter alone. In some
situations the sample can be examined using X-rays of two different energies
(frequencies) and the images combined to separate phases. In general, this
method cannot separate touching crystals of the same mineral. This is because
m does not depend on direction in an anisotropic mineral. This contrasts with
the optical properties of most common anisotropic minerals.
Sample size is limited by the attenuation of X-rays and the physical dimensions of the sample chamber. Many instruments can handle samples up to
40 cm long. Spatial resolution is a function of sample size and the number of
pixels in the image (typically less than 1000 1000 pixels). Resolutions are
commonly 0.1–0.2 mm for centimetric to decimetric size samples. Larger
samples necessarily have a lower resolution.
Recently, synchrotron radiation has been used for micro-tomography (Song
et al., 2001, Cloetens et al., 2002, Ikeda et al., 2004). This has a wide range of
wavelengths and is very intense so that it can be focused in a small volume. It
has also been combined with X-ray fluorescence computed tomography to
give a 3-D compositional map (Lemelle et al., 2004). Although precision is low,
this method shows promise for some materials.
2.2.4 Magnetic resonance imaging
Another 3-D method is magnetic resonance imaging. This is used extensively
in medical applications, but has only recently been applied to textural studies
of rocks. Magnetic resonance generates images that largely reflect the hydrogen content of geological materials. So far this method has only been used in a
single textural study, where it was used to help visualise pore distributions in
carbonate rocks (Gingras et al., 2002). Clearly, it may be applied in other
studies of water-bearing rocks, especially if the resolution of the method can be
improved.
2.3 Extraction of grain parameters from data volumes
The digital 3-D analytical methods described above produce a 3-D image
commonly referred to as a data volume, or data brick. It is comprised of
14
General analytical methods
voxels, the volumetric equivalent of pixels in images. Two-dimensional image
analysis (see Section 2.6.3) is much more developed than analysis of threedimensional images. Many of the 2-D techniques can be extended to 3-D, but
the software for this is still in development.
Extraction of grain and textural parameters from data volumes comprises
three steps: classification, separation and measurement (Ketcham, 2005). The
classification process is commonly the most complex. Many data volumes are
essentially grey-scale images – that is there is only one value for each voxel.
Such images can be segmented by considering a window of acceptable values.
However, mineral phases are not always very regular, and more useful methods have been developed (Ketcham, 2005). For instance, the ‘seeded threshold’
filter initially accepts voxels within a range of grey-scale values. Each seed
object is then expanded by the addition of connected voxels that have a wider
range of grey-scale values. If the mineral phase is assumed to be spherical then
irregular groups of voxels may be simplified by substituting spheres with
volumes equal to that of the original crystal (Carlson et al., 1995).
Touching crystals or grains are not separated by most 3-D analytical methods; hence this must be done during data reduction. Voxel groups can be
examined individually and cut apart manually (Ketcham, 2005). An automatic
process was suggested by Proussevitch and Sahagian (2001); interconnected
voxel clusters are ‘peeled’ or eroded until the individual objects are separated,
and finally the crystal centres are defined. The crystals are then rebuilt using an
assumed shape, such as a sphere or equidimensional polyhedron. Another
approach is to use the ‘watershed’ algorithm: the acceptable range of voxels in
a group is reduced until the group separates into distinct objects. The voxel
group is then rebuilt from these centres (Ketcham, 2005).
Measurement of the dimensions of separated groups of voxels is conceptually easy, but has not been facilitated by current software. However, new
developments may ease this problem (Ketcham, 2005).
2.4 Destructive partial analytical methods
Some aspects of rock textures can be determined by dismantling the rock and
measuring the dimensions of the separated grains. This avoids problems of
interpreting the grain parameters from sections, but the position and orientation of the grains are lost. In addition, grains with convoluted shapes may not
be easily separated intact from their matrix and it is difficult to know how to
deal with edge effects. Such analytical methods enables the use of much smaller
samples as many more crystals are encountered in volume compared to those
intersected in a section. However, the minimum crystal size that can be
2.4 Destructive partial analytical methods
15
consistently recognised must be clearly established (and also the maximum size
in rare cases).
2.4.1 Sample disaggregation
Grains can be separated from a rock if it can be readily disaggregated. In some
volcanoes explosive eruption processes may separate crystals mechanically.
However, it is unlikely that this process will be truly unbiased in terms of size
or shape. Other volcanic rocks may be so weak that they can be separated
mechanically with little damage to the crystals (also see sample solution
methods below). Dunbar et al., (1994) did this for bombs from Mt Erebus,
Antarctica, with good results. However, they did have to make a correction for
some broken crystals.
Well-lithified rocks can be disaggregated using more energetic methods. In
electric pulse disintegration a voltage of 100 kV is applied to a rock sample
sitting in a water bath (Rudashevsky et al., 1995). The rock explodes, separating crystals along grain boundaries. The specialised nature of the equipment
has limited the use of this method so far. Finally, it is possible to dissect a rock
crystal by crystal using a hammer and chisel. Kretz (1993) used this method to
measure the position of each garnet crystal in a schist. He then reconstructed
the whole structure with balls and rods.
Crystals that are already fractured in the rock present special problems. If
one is interested in primary processes then one solution is to examine each
crystal individually and only retain those that are unbroken. This is time
consuming and may bias the sampling of the crystals, but it may enable
some studies that would otherwise be difficult (Gualda et al., 2004). Of course,
if the focus of the study is on the fracturing process then the crystals can be
measured easily.
2.4.2 Sample dissolution
Dissolution is another method for the separation of crystals from a rock. In
carbonate matrix rocks (carbonatites and marbles) some crystals can be separated by dissolution of the matrix using HCl or other acids. However, it should
be remembered that some silicates are also slightly soluble in HCl (e.g.
anorthite, olivine) and hence small crystals may be lost.
Similar dissolution methods are commonly used to separate diamonds from
kimberlite or other rocks for exploration purposes. The preferred method is
crushing followed by fusion with alkali flux at 550 8C. Commercially very large
samples are processed – up to 300 kg. Larger crystals may be broken during
16
General analytical methods
crushing that precedes dissolution. In that case the crush size is the maximum
crystal size that can be recognised. In some kimberlites most diamonds are
already broken, probably during emplacement; hence the size measured is that
of the fragments, not the original crystals. The smallest crystals may be lost by
solution or mechanically.
Dissolution methods are also used to extract crystals from silicic volcanic
rocks (Bindeman, 2003). The method works best for light, frothy pumice.
Hydrofluoric (HF), fluorosilicic (H2SiF6) or fluoroboric (HBF4) acid is used
to attack the glass. The acid is applied either until there is complete dissolution
of glass or until the glass is weakened by partial dissolution and the rock can be
easily crushed. Many silicate minerals are also attacked by the acid, but
generally much more slowly than the glass. The surfaces of feldspar crystals
are etched more than quartz, which can be used to distinguish these minerals.
Small crystals may dissolve, stick to the surfaces of the preparation equipment
or be retained on filters. However, good results have been obtained for zircon
and quartz in rhyolitic pumice (Bindeman, 2003).
Mixtures of different minerals extracted by dissolution can be separated by
density using heavy liquids (methylene iodide ¼ diiodomethane, bromoform ¼
tribromomethane, sodium polytungstate) or magnetically using a Frantz#
isodynamic separator (Hutchison, 1974). They can also be hand picked dry,
under alcohol (to reduce reflections) or in immersion oil. In the latter case the
refractive index of the oil can be matched to that of the matrix glass or another
mineral, to make it less visible.
2.5 Surface and section analytical methods
2.5.1 Surface preparation techniques and artefacts
Most quantitative textural studies of rocks start with the preparation of an
artificial flat surface: natural fracture surfaces cannot generally give quantitative results. Commonly, the first step is sawing a rock sample with a diamondimpregnated circular or wire saw. The rough surface can then be flattened on a
lap (rotating wheel) with abrasive paste. The surface is polished with progressively finer grained abrasives until the necessary degree of flatness has been
achieved. A number of problems and artefacts are commonly encountered:
they should be recognised so that they will not be misinterpreted. This is
especially a problem with automatic analysis systems.
*
Scratches: These are not always easy to remove. They can be a problem if the rock is
comprised of minerals with variable hardness. Surface treatments like etching can
enhance small scratches.
2.5 Surface and section analytical methods
*
*
*
17
Occluded materials: Grains of the grinding material can be pressed into softer
minerals, or can be caught along grain boundaries or cracks in the sample.
Pull-outs: Brittle minerals with well-developed cleavages can fracture close to the
surface making small pits.
Rounding and surface relief: In polymineralic rocks harder minerals will resist
abrasion and will tend to be higher in the final polished surface.
The relief of the surface can be increased by etching. This is commonly used for
Nomarski microscopy (see Section 2.5.3.3) but has also been applied in other
studies. Herwegh (2000) developed a two-stage etching for calcite: the surface
is first immersed in dilute HCl, followed by dilute acetic acid. The surface relief
was then examined with a scanning electron microscope; however, Nomarski
microscopy could also be applied.
2.5.2 Electronic and associated analytical methods
Rock surfaces can be examined using a beam of electrons and the most
common instrument for this is the scanning electron microscope (SEM; Reed,
1996). The sample is placed in a vacuum chamber and a beam of electrons is
scanned across the surface. Interaction of the beam with the material produces
electrons, X-rays and light photons that are detected and measured. The
resolution of the different images is variable, but is theoretically smaller than
for optical measurements as the wavelength of electrons is much smaller than
that of light.
Samples must be flat and highly polished otherwise it is the relief that will be
imaged instead of the composition. Solid samples or polished thin sections can be
used, with the only physical limitation being the size of the sample chamber,
typically less than 10 cm. Samples are commonly coated with carbon or metal to
make them conducting and to reduce charging of the surface (Reed, 1996). In
some instruments the sample chamber is kept at a higher pressure than the
electron gun and hence no coating is necessary (ESEM – environmental scanning
electron microscope). The magnification can be varied enormously, making this
technique very useful over a wide range of sample sizes from 0.1 mm to 1 mm.
The electron microprobe (EMP) is another instrument that uses electrons to
analyse materials. It is mechanically very similar to the SEM, but has been
optimised for different measurements. SEMs are designed for observations at
different magnification scales, but the sample cannot be viewed optically.
EMPs commonly have a magnification fixed to that of the associated optical
system. In addition EMPs are optimised for quantitative chemical analysis.
Recently, there has been a convergence between these two instruments, but
SEMs are still cheaper than EMPs for both purchase and use.
18
General analytical methods
2.5.2.1 Backscattered electron images
When a beam of electrons strikes a surface some electrons will pass close to
the nucleus of the atoms and will be scattered by the positive charge of
the protons. If the electron beam is approximately normal to the mineral
surface (‘flat scanning’) then the number of ‘backscattered electrons’ (BSE)
emitted by any part of a rock is proportional to the mean atomic number of the
mineral, Z:
P
Zj Nj Rj
Z ¼ P
Zj Rj
where Zj ¼ atomic number; Nj ¼ atomic weight and Rj number of atoms in the
formula of element j. If the beam is strongly inclined to the surface then other
factors come into play (see orientation contrast imaging below).
The spatial resolution of BSE images is limited to about 0.1 mm as the
electrons are produced within a relatively large volume. The atomic number
difference that can be distinguished in a BSE image also decreases with
increasing atomic number (Reed, 1996): At Z ¼ 10 u (e.g. quartz,
feldspars, Table 2.1) it is 0.1 u and at Z ¼ 30 u (e.g. Cu-sulphides) it is 0.5 u.
However, the overlap of Z ranges of potassium feldspar, plagioclase and
quartz is much more of a problem, hence supplemental information, such as
X-ray maps may also be necessary. BSE images are most useful for small
crystals that have significantly different Z from other minerals and the
groundmass.
for various
Table 2.1 Mean atomic numbers (Z)
minerals. These are derived from actual analyses in
the MinIdent-Win 3 mineral property database
(see Appendix).
Mineral
Mean atomic number (Z)
quartz
albite
anorthite
orthoclase
fayalite
forsterite
orthopyroxene
10.8
10.8
11.9
11.7
18.3
11.4
13.8
2.5 Surface and section analytical methods
19
2.5.2.2 X-ray maps: EMP, SEM and Micro-XRF
X-ray maps are images in which the intensity (pixel value) is related to the
composition of the surface. They are created from analysis of secondary
X-rays produced when electrons strike a surface with sufficient energy
(Reed, 1996). The energy (or wavelength) of some of these X-rays are characteristic of the atomic number of the atoms in the target and hence can be
used to determine the elemental concentration. The quantity of X-rays produced per unit mass decreases considerably with atomic number, as does the
sensitivity of the detector and window systems.
All EMPs and most SEMs have the X-ray detectors needed to produce
X-ray maps. There are two types of X-ray detector. Energy-dispersive detectors use a silicon or germanium crystal and can measure many different X-ray
energies simultaneously. Most are not sensitive to elements lighter than
sodium and the resolution is not always sufficient to separate adjacent peaks
produced by different elements. Wavelength-dispersive detectors use a crystal
to diffract the X-rays produced by the sample. The angle between the detector,
sample and the crystal is varied to select different X-ray wavelengths and hence
only one element can be measured at a time. The resolution and sensitivity of
this detector are greater than those of the energy-dispersive detectors and hence
they are more sensitive and precise.
If an electron beam is scanned (rastered) across a sample then the X-rays
emitted at each point can be filtered for each element. This can be assembled to
give an X-ray map of the sample for each element. For larger samples the beam
can remain stationary and the sample driven mechanically to give the same
effect. Such images overcome some of the problems associated with BSE
images, in that minerals such as quartz and feldspars are clearly distinguished.
However, production of X-ray maps is time-consuming and hence costly. In
addition, the resolution of the images is commonly poor compared to BSE
images as the X-rays are much less intense.
Secondary X-rays are also generated when a beam of primary X-rays strikes
a surface. This is called X-ray fluorescence (XRF) and is commonly used for
the chemical analysis of powder samples. Recent developments in X-ray optics
have enabled the beam size to be reduced to 50 or even 10 mm in a micro-XRF
instrument. While this resolution is much less than that of an SEM it is
sufficient for many studies. The X-rays are generally analysed with an
energy-dispersive detector, as for an SEM. There are many advantages to
this technique: the equipment is much cheaper and the analyses faster. It can
operate without a vacuum and there are no electronic charging effects, as
X-ray photons are neutral.
20
General analytical methods
2.5.2.3 Cathodoluminescence
Cathodoluminescence is the emission of light by a crystal in response to
electron bombardment (Pagel, 2000). This effect is only seen in some minerals, but these include such common species as feldspars, calcite, zircon and
quartz. The intensity and colour of the light are highly variable and commonly depend on the concentration of trace elements called activators and
the density of lattice defects. Hence, cathodoluminescence can be used to
distinguish grains or parts of grains with different growth histories. It is used
extensively to examine the petrology of sedimentary rocks, but has been less
exploited in textural studies of metamorphic and igneous rocks (e.g. Titkov
et al., 2002).
Cathodoluminescence is most commonly measured with a luminoscope: a
special instrument that is attached to a regular petrographic microscope. It can
also be observed with the optical system of an EMP. The most sensitive
method is to use a special light detector in an SEM, but this only records the
intensity and not the colour of the light.
2.5.2.4 Orientation contrast imaging
Under normal ‘flat scanning’ mode the mean atomic number of the crystal (Z)
controls most of the variation in BSE intensity. However, if the electron beam
hits the sample obliquely then the crystallographic orientation of the crystal
becomes important because electrons are channelled into and out of the
crystals along lattice planes (Figure 2.4; Prior et al., 1999). If these are parallel
to the electron beam and/or direction of the detector the electron intensity will
be enhanced. This effect is exploited in orientation contrast imaging (OC). In
flat scanning the OC effect is about ten times less important than Z effects.
However, if the sample is tilted at 708 to the beam direction then OC dominates
(Figure 2.4).
Surface preparation is particularly important in OC. Normal polishing
(sufficient for BSE images) disturbs the crystal lattice near to the surface and
inhibits the OC effect. Therefore, final polishing must be done using special
techniques, such as colloidal silica (Xie et al., 2003), chemical–mechanical
polishing or etching (Prior et al., 1999).
This method may be especially useful for imaging touching cubic (isometric)
crystals. Such minerals are optically isotropic and hence cannot be separated in
reflected and transmitted light. However, OC varies with the orientation of the
lattice planes; hence the crystals will have different intensities and grain
boundaries can be easily identified.
2.5 Surface and section analytical methods
21
Electron
beam
Crystal 1
Channelled
electron
beams
Crystal 2
Figure 2.4 Orientation contrast imaging. The lattices of two crystals with
different orientations are indicated by the arrays of points. In this example
the electron beam is vertical and the sample surface is at an angle of about
70 degrees. Electrons are channelled into and out of the crystals along
lattice planes.
2.5.3 Optical analytical methods
Examination of rocks in transmitted and reflected light has a long history and
still remains a very cost-effective analytical method. Indeed, it is probably the
most commonly used method in quantitative textural analysis. The wide range
in the optical properties of minerals, especially their common anisotropy,
makes it easy to distinguish individual crystals, even when they touch. Hence,
it is the method of choice for minerals that are especially abundant in a rock.
The quality of images is very important for successful quantification. Many
older microscopes have high-quality dedicated (and commonly antiquated)
film cameras. Individual photographs taken on 35 mm film can sometimes be
enlarged to 35 cm without loss of resolution. The prints can then be scanned
for digital processing. Small ‘domestic’ digital cameras commonly have complex zoom lenses which can have considerable flare and chromatic aberration.
More expensive dedicated digital cameras have much simpler and better lenses
and greater resolution.
22
General analytical methods
2.5.3.1 Transmitted light
Rocks are examined in transmitted light using thin sections. A standard thin
section is 30 mm thick and measures 23 cm. Larger thin sections, 3 7 cm, are
also readily available, as are thinner-than-normal sections. Polished thin sections that can also be used in reflected light are generally only available in the
smaller size. Identification of minerals by optical methods is complex and is
covered by basic texts, such as Nesse (1986).
Thin sections are usually examined in plane-polarised light and under
crossed polarisers with a petrographic microscope. Normal petrographic
microscopes equipped with film or digital cameras can yield good photographs
of regions up to 2 mm. Unusual low-power objectives (1) can image larger
areas, but special stages are needed and illumination is commonly uneven.
Larger images can be assembled by taping together photographic prints into a
mosaic or by electronically stitching digital images (see Appendix).
It is difficult to get an image of a whole thin section with a regular petrographic microscope. Some specialised zoom microscopes can cover a whole
regular thin section with as few as 10 images. Other solutions found are a film
scanner (Tarquini & Armienti, 2001), a slide-copying apparatus or camera
attachment, or a light bench. Some systems use a flash so that vibration
problems are reduced. It is also possible to use a document scanner that can
operate with transmitted light on a thin section. This works well for a direct
image or with plane-polarised light produced by a single Polaroid sheet.
However, if two Polaroid sheets are sandwiched on either side of the section
to make a cross-polarised image then there is not always enough light to yield a
good image.
The lower limit of resolution is commonly close to the thickness of the
section, 30 mm. Although the physics of light indicates that a higher resolution
is possible, it is rarely approached in most studies because of the difficulty of
observing small crystals in thin section, except where the matrix is lightly
coloured. It should also be remembered that small crystals enclosed in the
thin section are seen in projection whereas larger crystals are seen in section.
This necessitates different mathematical procedures to calculate the 3-D
textural parameters (see Chapter 3).
A multitude of colour images are possible from a single thin section: in
plane-polarised light pleochroic minerals will have different colours according
to the orientations of the polarisers. Similarly, when examined with crossed
polarisers crystals will show different colours and extinctions according to
the orientation of the polarisers. Most types of automatic image analysis
cannot cope with more than a single image, however, computer-integrated
2.5 Surface and section analytical methods
23
polarisation microscopy methods are designed to exploit this aspect of optical
mineralogy (see Section 2.6.4; Heilbronner & Pauli, 1993, Fueten, 1997).
In cross-polarised light some crystals will generally be in extinction for any
orientation of the section. This can be a problem for the measurement of
crystals using simple observations of a single cross-polarised image. A littleused optical technique, the ‘Benford plate’ can eliminate this problem (Craig,
1961). The name is a misnomer as the set-up consists of two matched quarter
wave plates (retardation ¼ 132 nm). One is placed below the thin section,
oriented at 908 to an accessory plate holder. The other is inserted in the normal
position for an accessory plate. With the Benford plates installed, the birefringent colour of crystals is identical to that seen at the 458 position without the
plates. However, the birefringence colour, and its intensity, do not change with
rotation of the stage; that is, extinction no longer occurs. Clearly, this can be
very useful as all crystals can be observed with the stage in a single orientation.
However, the intensity contrast between grains may be less than that normally
observed. This image is identical to the maximum birefringent image obtained
in computer-integrated polarisation microscopy (see Section 2.6.4).
Thin sections are generally observed orthogonal to the plane of the section,
but in some situations it may be useful to remove this constraint. The universal
stage is an accessory for a standard petrographic microscope that enables
tilting of the section with respect to the direction of observation. Hence,
crystals and their boundaries can be examined from different orientations. It
is described in more detail in Section 5.5.1.
The infrared absorption of mineral sections can also be examined, generally
by using a high resolution infrared spectrometer (e.g. Ihinger & Zink, 2000).
A section thicker than normal (800 mm) is polished on both sides and the
spectra of 100 mm spots collected. Each spectrum is filtered for absorbance in
a range of wavelengths that correspond to resonance of specific chemical
entities and is used to build an absorbance map of the section. This technique
has been used to examine the distribution of hydrogen in quartz.
2.5.3.2 Reflected light
Rocks are examined in reflected light using polished thin sections or blocks
(Cabri & Vaughan, 1998). Examination of polished sections in orthogonally
reflected light can complement studies in transmitted light or be done independently. Many research microscopes are set up so that they can be used for
both optical techniques. The variation in optical properties in reflected light
is generally more limited than that in transmitted light, but the method is
especially important for opaque minerals. The resolution of the method is not
limited by scattering of light within the material and hence features as small as
24
General analytical methods
1 mm can be readily observed. However, there is no good way to get images of
areas larger than can be viewed with the normal microscope lenses or
assembled from image mosaics. For instance, a document scanner does not
yield very good images because the light is not orthogonal to the surface.
However, it may be useful if there are very large differences in reflectivity.
Although surfaces may be viewed in both plane-polarised light and under
crossed polarisers, the former is used most commonly for imaging surfaces. As
for transmitted light, anisotropic minerals may have different reflectances
(colours) for different orientations, but the effect is not large for most minerals. Hence, minerals are usually imaged in plane-polarised or unpolarised
light. In this situation the most important parameter is the reflectance of the
mineral. Reflectances have been recently tabulated for all new minerals, at
20 nm intervals from 400 to 700 nm, but values at four standard wavelengths
are generally used (470, 546, 589 and 650 nm). Most transparent minerals, such
as quartz and feldspars have reflectances in the range 5–10%. Other minerals
are much higher: oxide minerals 12–30%; sulphides 12–60% and metals
50–100%. Surfaces may also be etched or stained to bring out the contrast
between different minerals, sub-grain boundaries and crystal defects (see
Section 2.5.3.3 and review by Wegner & Christie, 1985).
Minerals can be imaged electronically by standard charge-coupled device
(CCD) cameras. However, such cameras are designed to mimic the human eye
and hence record light in three wide spectral bands. A better approach is to use
narrow (10 nm) bandwidth filters (Pirard, 2004). Such a system needs to be
carefully calibrated, but can give much better resolution of mineral phases,
especially those with reflectances greater than 5%. This method can distinguish mineral pairs that are problematic in BSE (e.g. chalcopyrite/pentlandite)
or X-ray maps alone (e.g. hematite/magnetite/goethite).
2.5.3.3 Nomarski (DIC) microscopy
A very useful optical technique for imaging surfaces is differential interference
contrast (DIC) or Nomarski microscopy. It has long been used by metallurgists to produce detailed images of polished surfaces, and has also been used by
petrologists to examine mineral and rock textures (Anderson, 1983, Pearce
et al., 1987, Pearce & Clark, 1989). It can reveal both the exterior shape of
crystals and their internal structure, and can be useful for distinguishing
crystals from a glassy matrix. It can be applied to thin sections and hence is
complementary to normal reflected and transmitted light methods.
Nomarski microscopy is an optical technique for imaging the micro-relief of
surfaces (<0.5 mm). A light beam is split by a prism; one beam is reflected off
the surface and allowed to interfere with the reference beam. Hence the method
2.5 Surface and section analytical methods
25
Table 2.2 Etchants used for Nomarski examination of minerals. All samples
should be neutralised with Na2CO3 after etching so that degassing of HF does not
etch the objective lenses. A much longer list of possible etchants and further
details of the methods are available in Wegner and Christie (1985).
Mineral
Etchant
Notes and reference
Plagioclase
Fluoroboric acid (HBF4); 2–3
minutes at room temperature
Concentrated HCl; 10–20 minutes
at 45 8C
Concentrated HF; 2–4 minutes
at room temperature
(Anderson, 1983)
Olivine
Clinopyroxene
(Clark et al., 1986)
HF attacks plagioclase and
olivine vigorously (Clark
et al., 1986)
is sensitive to surface relief of the order of the wavelength of light. The special
lenses and prisms that are needed are usually fitted to a specialised microscope,
which may be available in metallurgy laboratories.
The initial surface must be very well polished, with no remaining scratches.
This can be done in the same way as for normal polished sections. The surface
is then etched to develop relief (Table 2.2; Wegner & Christie, 1985). Etch
depth will depend on the nature of the mineral, the orientation of the section
and the density of crystalline defects. Finally, the surface may be coated with
carbon or metal using the same apparatus that is used for SEM studies. This
coating reduces interference from reflections beneath the surface of the
sample. However, it also reduces the contrast in reflectivity between different
minerals and glass. If polished thin sections are used then the technique may be
complemented by examination with transmitted light, if the coating is not too
thick. This is useful if two anisotropic crystals touch; they may then be
distinguished by differences in optical orientation.
2.5.4 Slabs and outcrops
For rocks containing large crystals sawn slabs can be very useful. The
surface should be planar, but it is not usually necessary to have a perfectly
polished surface, unless very small crystals will be examined. A sawn surface
can be ground flat with wet abrasives on a rotating lap (wheel) or with
waterproof (wet and dry) silicon carbide/oxide paper on a flat surface such
as a sheet of glass.
The surface can be etched or stained to enhance the contrast between the
different minerals. One of the best known stains is sodium cobaltinitrite which
26
General analytical methods
Table 2.3 Some staining procedures that can be used to help distinguish
minerals in slabs. Some of these techniques can also be used for staining thin
sections. Details of these and other procedures can be found in Hutchison (1974).
Minerals
Treatment
Colours and notes
K-feldspar
1) HF (49%) 30 seconds
2) Na3Co(NO2)6 Sodium
cobaltinitrite (freshly
prepared) 10 seconds
Amaranth red
1) HCl (1.5%) 10 seconds
2) Alizarin red
3) Potassium ferrocyanide
K4Fe(CN)6 3H2O
K-feldspar ¼ orange
Plagioclase ¼ white
Quartz ¼ grey
Quite resistant to abrasion
Red. Easily removed from surface
Calcite ¼ rose to red
Dolomite ¼ colourless
Plagioclase
Calcite and
dolomite
colours all potassium feldspar a deep orange. Other minerals can also be
stained (Table 2.3; Hutchison, 1974). It is also possible to stain thin sections.
Crystals in slabs can be measured manually in several ways: (1) The simplest
is with a ruler and protractor (see below). (2) Mineral outlines can be traced
onto a transparent overlay and subsequently scanned and analysed automatically. (3) Slabs can be placed on a document scanner and the images analysed
automatically or manually. Very large polished rock panels, as used for building facing, can provide material transitional in scale between slabs and
outcrops.
Very large crystals must be examined in outcrop, especially those with low
number densities. Crystals can be measured directly with a ruler and protractor. Photographs of outcrops can also be used, but not generally so successfully – what is clear in the field can be confusing on a photograph at a later
date. The method is most easily applied to flat surfaces naturally polished by
glacial action (Higgins, 1999). However, it can also be used on surfaces
smoothed by rivers or the sea. In some cases dimensional stone quarries may
yield sufficiently smooth surfaces.
2.6 Extraction of textural parameters from images
Some of the analytical methods described above yield quantitative textural
data directly. However, most produce images that must be measured to extract
various textural parameters. Manual image analysis methods demand a large
amount of individual attention and judgement and user training is very
important. Such methods generally yield high-quality data directly, but are
2.6 Extraction of textural parameters from images
27
very time-consuming. In addition, there is always the possibility of operator
bias, especially where the researcher does the measurement. Automatic image
analysis methods may take much time to set up, but can be faster if many,
similar samples need to be processed (see the general review in Russ, 1999).
The quality of the data is commonly not as high as that produced manually,
even though it may suffer less from operator bias. This may be balanced by a
greater number of crystals and samples measured.
The images produced by many analytical methods have too few pixels for
adequate precision. Several images can be combined together like a mosaic to
produce a larger image. This can be done electronically or by taping photographic prints together and scanning the mosaic.
2.6.1 Size limits of measurements
No matter what method is used to acquire quantitative textural data, the minimum size of crystal that can be recognised and measured must be established. If
no data are listed for crystals smaller than a certain size it is important to indicate
if this reflects an artefact of measurement or a real lack of crystals in that size
range. Methods can be combined to cover a wider size range than would be
possible with a single method. For instance data from outcrop measurements can
be combined with data from slabs; thin sections can be measured at two different
scales; BSE images can be combined with thin section measurements. It is very
important to record and publish the details of the data-acquisition method and
steps taken to ensure adequate quality control.
The maximum size of crystal that can be precisely determined is generally
determined by the number of crystals in the largest interval, and not by the
physical limitations of the method. However, the maximum size must also be
specified where possible.
2.6.2 Manual image analysis
Crystals viewed with an optical microscope can be measured directly without
recording an image: the scale in the microscope eyepiece can be used to
measure intersection length and width, and the crystal rotated on the stage
to determine the orientation. This has the advantage that different magnifications and orientations of the stage can be used to identify the crystal; hence a
wide range of crystal sizes can be measured. It is also useful for rocks that have
a low number of crystals per unit volume (number density). However, it is
difficult to keep track of which crystals have been measured, and the use of a
photograph is advised, even if only to check off that a crystal has been
28
General analytical methods
measured. A similar method can be used with an electron microscope (SEM),
with the same caveats.
Most of the techniques discussed above produce images: single images or
mosaics, on paper or in electronic form. These images can then be measured
manually in several different ways: (1) the intersection length, intersection
width, long-axis orientation and centroid position can be determined using a
ruler and protractor; (2) the crystal outlines may be traced onto a transparent
overlay, which is then scanned and analysed automatically as discussed below;
(3) digital images can be imported into a vector drawing program (e.g.
CorelDrawTM) and the crystal outlines traced by drawing a polygon with the
mouse; (4) photographs can be put on a digitising tablet and the outline of each
crystal in the photograph traced using the digitising puck. A number of
programs can be used to control the digitising tablet (see Appendix) and
reduce the raw positional data to textural parameters. In each case the actual
outline of the crystal in the thin section should be verified with a polarising
microscope.
2.6.3 Automatic image analysis
Automatic image analysis commonly involves many different steps
(Figure 2.5). The goal of image processing is to segment the original image
into a classified image of the mineral of interest: a wide range in pixel values in
one or many channels of the original data image are reduced to a small number
of different pixel values, one for each phase of interest. Clearly, it may not be
easy to classify some pixels. This commonly occurs where a pixel traverses the
contact between two minerals. Sub-pixel grains and unknown minerals can
also cause problems. It must be decided if it is really necessary to classify all
pixels or if unclassified pixels can be accepted.
A number of successful image processing procedures have been published
but many are specific to particular proprietary image processing software (e.g.
Perring et al., 2004). The expense and limited distribution of these systems has
led many researchers to use general purpose image software (i.e. Adobe
PhotoshopTM) and powerful open-source, free software (e.g. NIHImage/
ImageJ). Here, I have tried to present an approach that can be applied more
generally. I do not want to enter into the specifics of program usage, as this will
be outmoded before the book is published.
2.6.3.1 Grey-scale images
Some analytical methods produce images with a single variable, usually called
grey-scale images: for instance, the amount of backscattered electrons or the
2.6 Extraction of textural parameters from images
Grey-scale image
(backscattered electron
reflected light, tracing
of photographs, etc.)
Sharpening
Smoothing
Trimming
29
Multichannel image
(Transmitted light,
X-ray mapping, etc.)
Sharpening
Smoothing
Trimming
Contrast
Brightness
Contrast
Brightness
RGB Z HSI
Channel separation
Processed
grey-scale image
Thresholding
Classified image
Processed
multichannel image
/
ng
mi
m
gra et
pro ral n
d
c
i
an
t
ne neu ning tion
e
a
i
G
Tra ssific
cla
Training and
classification
Principal component
analysis
False colour image
De-noising
Manually separate
touching crystals
Final classified
image
Measurement
(NIHImage, etc.)
Numerical
intersection data
(length, width, area,
orientation, position . . . )
Figure 2.5 Flow diagram for automatic image analysis of grey-scale (single
channel) or multichannel images. This diagram does not show all possible
strategies, just the most common routes.
overall reflectance in reflected light. Grey-scale images can be smoothed,
sharpened and trimmed to improve recognition of individual mineral grains.
Such manipulations can be done using readily available general purpose,
image-processing software (GPIP; see Appendix; e.g. Adobe PhotoshopTM
or Corel PhotopaintTM) but dedicated programs commonly make it clearer
what has been done to the data and how it can be repeated (e.g. LViewProTM).
Processed grey-scale images are easily segmented by selecting upper and
lower thresholds (tones of grey) for each mineral by ‘training’: the values
30
General analytical methods
associated with a mineral are determined for a known crystal. Pixels that differ
considerably from those of the known minerals are best ignored, even if this
means that part of the image will be unclassified.
2.6.3.2 Multichannel images
Many analytical methods produce images that have more than one value
associated with each pixel. The image associated with each variable is called
a channel. For instance, optical images of minerals may be coloured, which is
generally expressed using the red–green–blue model that mimics the behaviour
of the human eye. These data comprise three channels. Another approach is to
use narrow (10 nm) bandwidth filters, which can yield many more colour
channels (Pirard, 2004). X-ray images of surfaces can have a channel associated with each element. Such multichannel data avail themselves to more
complex processing than grey-scale data. However, the first processing step is
the same as for grey-scale images: the image can be smoothed, sharpened and
trimmed, and the contrast and brightness adjusted.
Pixels in multichannel images must be classified, but clearly the process is
considerably more complex than for grey-scale images. There are many possible ways to treat such images and only a simplified approach will be discussed
here. More details are available in texts such as ‘The Image Processing
Handbook’ (Russ, 1999) or ‘Image Analysis in Earth Sciences’ (Petruk, 1989).
For many optical images the largest variation between different phases is the
brightness of the image, rather than the colour. This is expressed as a significant correlation between the signal in the red, green and blue channels of an
RGB image (Launeau et al., 1994, Thompson et al., 2001). That is, if a pixel
has an intense red component it is likely that it will also have intense green and
blue components. Such images may be usefully converted to other colour
systems using GPIP: a useful system is hue, saturation and intensity or brightness (HSI – HSB). Hue is the basic colour, defined by the dominant wavelength
of light; saturation is the degree to which the hue is diluted by white light;
and intensity or brightness is the overall brightness of the colour. There are
a number of other schemes that may be useful also – trial and error is
recommended.
The variation between the pixel values for multichannel images (e.g.
colours) can be clarified by principal component analysis (PCA; Launeau
et al., 1994). The objective of PCA is to decorrelate and concentrate the
information carried by many channels. It is equivalent to rotating the axes of
the multichannel data in multidimensional space so that the greatest variation
lies along the first axis (first component) and the next orthogonal axis (second
component) contains the greatest variation in the remaining data. What
31
2.6 Extraction of textural parameters from images
variation is left lies in the other components. The main disadvantage of this
method is that each sample will give axes with different orientations. One
solution is to calculate a new set of axes from one sample and apply to all other
samples in the same dataset.
The processed multichannel image can be separated into component channels and each thresholded like a grey-scale image to yield binary data using
GPIP. It is also possible to select manually the range of colours associated with
each mineral. This is done by selecting the colour values associated with the
mineral using GPIP (in some programs it is necessary to make a colour ‘mask’
and to save that as a separate image). Another strategy is to substitute another,
stronger colour (e.g. bright green) for the mineral in question and then separate out that new colour. However, much trial and error is needed to ensure a
clean separation.
Multichannel data, either RGB, HSI or PCA-transformed components, can
be classified in a more elaborate way than just thresholding one channel.
Several approaches are possible but all start with training – that is measuring
the value of pixels that lie within a known mineral. In the parallelepiped
method the upper and lower thresholds are established for each channel,
giving a parallelepiped volume for each mineral phase (Figure 2.6a; Launeau
et al., 1994). Pixels that fall outside these boxes or in the intersection of two
boxes are unclassified. A better method is to define spheres rather than boxes
(Figure 2.6b). Where two spheres intersect the contact is planar, like a soap
bubble, which eliminates unclassified pixels that lie in two boxes. Pixels outside
(c) Maximum likelihood
channel 2
Unclassified
(b) Sphere
channel 2
channel 2
(a) Parallelepiped
Mineral A
Mineral B
Unknown mineral
channel 1
channel 1
channel 1
Figure 2.6 Simplified classification of pixels for two channels of multichannel
data. (a) Parallelepiped method. Pixels within the parallelepipeds are classified,
but those in the area of overlap are unclassified. All other pixels are
unclassified. (b) Sphere method. Pixels within spheres are classified. Outside
the sphere may remain unclassified or may be classified according to the
distance to the nearest sphere centroid. (c) Maximum likelihood method. All
points are classified according to their distance to the nearest ellipsoid centroid,
normalised by the ellipsoid radius in that direction.
32
General analytical methods
the spheres may be left unclassified or assigned to the sphere with the nearest
centroid. Marschallinger (1997) found that the best method was the maximum
likelihood classifier (discriminate-function analysis; Figure 2.6c). Results from
the training are used to construct ellipsoids instead of spheres. The shape and
orientation of the ellipsoid approximates that of the original ‘cloud’ of pixels.
A pixel is classified according to the distance to the centres of the ellipsoids
divided by the radius of the ellipsoid in that direction. This method is available
in dedicated image analysis programs (see Appendix).
Once the grains have been separated parameters can be evaluated for each
entire grain and used to classify the grains into different mineral phases. In
addition to the parameters mentioned above, it is also possible to quantify the
spatial variations in pixel values within each grain. This is referred to as
‘texture’, although this term clearly has a wider meaning in this text. So far
only the intensity component of the image has been used to evaluate four
standard textural parameters called contrast, entropy, energy and homogeneity (Thompson et al., 2001). These all relate to the way in which the intensity
varies between adjacent pixels. All these parameters are evaluated for each
grain and the results classified using neural networks (Thompson et al., 2001)
and genetic programming (Ross et al., 2001). Both give similar results, but
neural networks appear to be easier to use and more flexible.
2.6.3.3 Final processing of classified images
The resulting classified images can be manipulated easily in GPIP to remove
individual pixels or small groups of pixels, referred to as ‘noise’. Touching
grains in the classified mineral image can be separated manually by using GPIP
also: a line is drawn to separate the grains. If there are many overlapping grains
then an automatic procedure may be preferable. These methods usually require
that the crystal shape is known. The contours of the grains may be eroded by
removing the exterior pixels until the grains are separated. The lost pixels are
then added by dilating the grain back to its original size (Russ, 1999). Another
choice is ‘watershed’ algorithm (see Section 2.3). Finally, the sharp re-entrants
between touching grains can be detected and the outline segmented at this point
(van den Berg et al., 2002). Success rates vary from 50% to 60%, so this method
should not be used for samples with large numbers of overlapping grains.
2.6.4 Computer-integrated polarisation microscopy
All automatic processing of transmitted light images currently uses only
colours for classification of particles. However, an experienced observer uses
other parameters: for example the colour changes produced during rotation of
2.7 Calculation of 3-D data from 2-D observations
33
the stage under cross polarisers. These features may be examined using
computer-integrated polarisation microscopy (Heilbronner & Pauli, 1993,
Fueten, 1997). In this system of microscopy the polariser and analyser rotate
together, but the section, and hence the image stays still. Up to 200 images are
recorded for different orientations of cross-polarised light. These actual
images of the section can be combined and processed to give a series of
synthetic images, such as the maximum birefringence (Max) and angle of
extinction (Phi). Both Max and Phi give information about the lattice preferred orientation of crystals (Heilbronner & Pauli, 1993, Fueten &
Goodchild, 2001). They can also be used to segment the section into separate
crystals (Goodchild & Fueten, 1998, Zhou et al., 2004a, Zhou et al., 2004b).
An image similar to the maximum birefringence image can be obtained more
simply using the Benford plate technique (see Section 2.5.3.1).
2.6.5 Classified image measurement
Grain outlines in classified images produced by image analysis may be measured to determine their geometric characteristics. Parameters commonly
measured include position of centroid (centre of outline), length and width
or major and minor axes of a best-fit ellipse, perimeter, area and orientation of
long axis. Many programs can be used for this but the most popular is
NIHImage/ImageJ (see Appendix).
2.7 Calculation of three-dimensional data from
two-dimensional observations
2.7.1 Stereology
The problem of conversion of two-dimensional section data to threedimensional textural parameters is not simple for objects more geometrically
complex than a sphere and there is commonly no unique solution for real data
(see below). This subject is treated in a branch of mathematics called stereology.
There are a number of reviews and books on the subject, mostly aimed at the
biological sciences (e.g., Underwood, 1970, Royet, 1991, Howard & Reed,
1998). Particular aspects of stereology will be discussed in the relevant chapters.
2.7.2 Stereologically exact ‘global parameters’
There are a number of very useful relationships that can be used to calculate
various 3-D ‘global parameters’ from intersection measurements without any
34
General analytical methods
knowledge of the distribution of the phase. That is to say these relationships
are correct for all grain sizes, shapes and orientations, provided representative
sampling is done (Underwood, 1970, Exner, 2004). It is not even necessary to
separate individual crystals or grains. These values are termed stereologically
exact, unbiased or accessible parameters. They are the equivalent of mean
values, in that they only describe a limited aspect of the texture. Other textural
parameters are termed biased or inaccessible, but they are still measurable in
some materials, if assumptions are made about the crystal shape and fabric,
and will be described in later chapters (Underwood, 1970, Howard & Reed,
1998, Brandon & Kaplan, 1999).
The most widely used global parameter is the volumetric fraction of the
phase, VV. The subscript indicates that the volume of phase, V, is normalised
to unit volume, V. Delesse (1847) showed that the total area of a phase, A, in a
representative section normalised to unit area, AA is equal to VV. It may seem
counter-intuitive, but the fabric of the rock and the orientation of the section
are not important – any plane will give the same area fraction, which is equal to
VV. Crystal intersection areas are commonly measured at the same time as the
intersection length and width of the crystals and they can be used to calculate
the volumetric phase abundance. In addition, the fraction of a random line
that intersects the phase, LL is also equal to VV. However, many lines must be
measured to give adequate precision. Finally, the fraction of random or
equally spaced points that lie on the phase, PP is also equal to VV. This is the
basis of point counting. Hence
VV ¼ AA ¼ LL ¼ PP
Another common global parameter is the interface density, which is
the total surface area of grains, S, in a unit volume, SV. A test line or circle
is drawn on the section and the number of intersections with the surface,
P, counted (Figure 2.7). This value is divided by the test line length to
give PL.
Alternatively the number of grains intercepted, N, may be counted. This is
also divided by the length of the test line to give NL. Then
SV ¼ 2PL ¼ 4NL
If there are several phases in a material, then the ‘contiguity’ of phase a, Caa is
the proportion of the interface area shared between like grains. It is determined
from the number of test points with aa boundaries, PL(aa)
Caa ¼ PLðaaÞ=PLðTotalÞ
2.8 Verification of theoretical parameter distributions
35
Test line
length
SV = 2PL
Surface intersection counts
4
1 23
1
6
5
2
8
7
4
3
Grain intersection counts
SV = 4NL
Figure 2.7 Determination of the total surface area per unit volume, SV from
the number of surfaces or crystals intersected. Here, eight surfaces are
intersected which gives SV ¼ 2 8/length or four crystals are intersected
which gives SV ¼ 4 4/length.
Similarly the ‘neighbourhood’ is defined as the proportion of interface area
shared by two phases a and b, Cab. It is determined from the number of test
points with ab boundaries, PL(ab)
Cab ¼ PLðabÞ =PLðTotalÞ
Other parameters are the mean linear size of grains, L
L ¼ VV=PL
VV may be determined from area-based measurements or point counting. It is
also possible to determine the mean linear distance between grain centres, D
¼ ð1 VV Þ=PL
D
The degree of orientation of grains can be determined from the ratio of the
number of intersections in orthogonal directions. This is the intercept method,
which is discussed in Section 5.4.3.2.
2.8 Verification of theoretical parameter distributions
In natural systems, such as rocks, textural parameters never follow exactly a
prescribed distribution law. However, petrologists commonly want to know
if a distribution may be approximated by a distribution law, so that the
36
General analytical methods
parameters of that law can be used as a substitute for the true distribution.
Most of the diagrams in this book are designed specifically for various theoretical distributions. Graphical methods are commonly favoured by geologists
to show how well a real distribution corresponds to a theoretical distribution.
However, there are a number of mathematical parameters that can also be
used to test a hypothesis (Swan & Sandilands, 1995).
A hypothesis is tested using a ‘null hypothesis’: this is the idea that the data
do not differ significantly from that produced by random processes (e.g. Swan &
Sandilands, 1995). This hypothesis must be tested using a test statistic. There is
a level of significance associated with this hypothesis, commonly 5% or 10%.
This means that there is only a 5% chance that the distribution seen occurred
by random processes and hence has no significance.
Probabilities can be interpreted as a measure of the times an outcome will
occur for a large number of repeated experiments. However, a more useful
approach considers probabilities as just an expression of the confidence that a
model or theory is correct. When a result is said to be significant or not
significant the latter approach is being used.
2.8.1 Chi-squared test
The most widely applied test is the chi-squared (2) test. This is applied to
discrete data points.
2
¼
X ðOj Ej Þ2
Ej
where j is the class interval number, Oj is the observed frequency of that class
and Ej is the expected frequency of that class. Each class should have at least
five observations. The significant values of this parameter depend on the
number of observations and the required significance level. Tables are available in standard statistical texts (e.g. Swan & Sandilands, 1995).
The chi-squared test is very susceptible to the number of classes (Swan &
Sandilands, 1995): The null hypothesis may be more commonly accepted for
larger numbers of classes. For a small number of classes failure to reject the
null hypothesis (i.e. the data conform to the model distribution) may be due to
broad and unimportant similarities between the data and the model (Swan &
Sandilands, 1995).
Many other statistical tests have been proposed and the reader should
consult more specialised texts (e.g. Swan & Sandilands, 1995).
2.8 Verification of theoretical parameter distributions
37
2.8.2 Straight line distributions
A number of graphs are designed to give a straight line if the textural parameter follows the theoretical distribution. In this case the data can be regressed
and the intercept and slope determined. The correlation coefficient r2 is
commonly used to quantify how closely the observed data correspond to a
straight line. A value of one indicates a perfect straight line. However, the r2
parameter alone is not very useful, as the significance of values will depend on
the number of points on the graph. Tables are available to assess the level
of significance in terms of the number of samples and the value of r2. A greater
problem is that the significance of r2 will depend on the uncertainty in each
point.
A much better assessment of the level of significance can be made if the
uncertainty associated with each point is known (Bevington & Robinson,
2003, Higgins, 2006). Ideally the contribution of each point should be weighted
for its uncertainty. A goodness-of-fit parameter, Q, can then be calculated that
determines the probability that the observed discrepancies between the
observed and expected values are due to chance fluctuations (Figure 2.8). A
very low value indicates that the discrepancies are unlikely to be due to chance
fluctuations. This means that either the model is wrong or the uncertainties
have been underestimated or the uncertainties are not normally distributed. If
Q is greater than 0.1 then the null hypothesis is rejected and the parameter may
(a) Less accurate data
Q = 0.1
reject null
hypothesis
(b) More accurate data
Q = 0.001
accept null
hypothesis
Figure 2.8 Effect of uncertainty on the significance of a distribution. Both
diagrams have the same distribution and hence the same value of r2.
(a) Measurements are not very precise and a straight line can be drawn
through the uncertainty bars of the points. The value of Q is high enough
that the null hypothesis can be rejected. (b) The same points have been
remeasured more accurately. A line cannot be drawn through the uncertainty
bars and Q is low. Hence, the null hypothesis is accepted and the points do not
follow the theoretical distribution.
38
General analytical methods
follow the model distribution. A value of Q > 0.001 may be acceptable,
because truly wrong models generally have very small values of Q. Values
higher than 0.1 do not indicate that the data have more validity. A value close
to one may indicate that the uncertainties have been overestimated.
2.9 Summary
Many different analytical methods can be used in textural quantification and it
is not always easy to choose the most appropriate method. There must always
be a balance between the analytical effort and the expected results: A lengthy
method may mean that few samples can be analysed. In some cases the project
may benefit from simpler measurements on a larger number of samples.
A common strategy for the choice of methods is:
1. Scale: ideally, a study should start with the identification of the size scale – e.g.
10 cm megacrysts are best examined on an outcrop, whereas microlites may need
electronic methods.
2. Parameters: the textural parameters that are needed should be specified – e.g.
crystal orientations are not preserved when a sample is dissolved.
3. Minerals: it is important to decide what minerals will be examined and if individual
crystals can be distinguished by the method.
4. Equipment: access to equipment needed for the method must be established, both
financially and logistically. If access is not possible then a different project or
approach must be chosen.
It should be remembered that if a problem is not tightly constrained then many
studies can be done with very simple instruments that are available in almost
all geology departments: petrographic microscope, camera, ruler, document
scanner and computer.
3
Grain and crystal sizes
3.1 Introduction
One of the most commonly quantified aspects of rock texture is the size
distribution of crystals, grains and, more rarely, blocks. Although the first
published study of crystal size in igneous rocks was in the early days of
petrology (Lane, 1898), the subject really only took off in 1988 after the
publication of seminal papers by Marsh and Cashman (Cashman & Marsh,
1988, Marsh, 1988b). In metamorphic petrology the study of size distributions
started a bit earlier with the study of garnets by Kretz (1966a). The importance
of grain size distribution in sedimentology completely outweighs that of other
branches of petrology and is covered in more detail in sedimentological texts
(e.g. Syvitski, 1991). Engineers have a considerable interest in the mechanical
and chemical properties of materials (Brandon & Kaplan, 1999): grain size and
composition play an important role in controlling these properties and hence
their quantification is very important (Jillavenkateas et al., 2001, Medley,
2002). Also of interest are pore and bubble sizes in rocks and these are
discussed separately in Chapter 7.
3.1.1 What are size and size distributions?
The size of a crystal or grain is a measure of the space occupied by the object. It
is a three-dimensional property and can be defined in several different ways,
according to the application of the parameter: if the shape of a crystal changes
during growth or solution then a volumetric measure of size may be useful.
Size based on volume is simple because every grain has a unique volume.
However, if we are interested in the growth or solution of crystals where the
shape is almost invariant then a linear measure of size is usually appropriate.
Hence, the change in size of a crystal can be expressed in terms of linear growth,
39
40
Grain and crystal sizes
mm per second, for example. In most geological applications a linear measure of
size is used and various definitions are discussed in Section 3.1.2. From here on
the term size refers to a linear measure of the space occupied by a grain.
Petrologists commonly describe the mean size of crystals or grains in a rock,
without clearly defining what they mean. The definition of size is discussed in
Section 3.1.2, but here I want to examine the significance of mean sizes. If a
rock contains grains smaller than the resolution limit of the analytical method
then the value of the mean size will depend on the resolution (Figure 3.1). In
this situation the mean can be defined only if we can describe a size function for
that unknown interval. In any case although the mean is a useful parameter, it
has only one dimension. Another one-dimensional parameter that is sometimes measured is the crystal or grain number density. This is defined as the
total number of crystals or grains per unit volume, although some authors use
this incorrectly for number density by area. Again, there is a problem if a rock
contains grains smaller than the resolution limit of the method. Both these
parameters are sub-sets of a more information-rich parameter, the crystal
size distribution (CSD), grain size distribution or particle size distribution.
This has two dimensions; size and crystal or grain number. It is defined
between upper and lower size limits. Outside these limits it is indeterminate,
but changing the measurement limits will not change the CSD values within
the limits (Figure 3.1).
Finally, some workers loosely use the terms unimodal and bimodal size
distributions. The number of modes depends clearly on the variable examined
and how it is expressed. For instance, the size variable may be expressed in
terms of length or volume of grains: this will clearly change the distribution. In
many ‘bimodal’ rocks there is actually only one mode in the CSD. What is
frequently meant by this term is that there are two populations of crystals.
However, this should be verified by other means, such as chemical or isotopic
composition, as single populations may appear visually to be multiple
populations.
3.1.2 Linear size definitions
The linear size of a grain has many different definitions. This is compounded
by a common confusion between true size in three dimensions and size in
intersections or projections of grains. Three-dimensional linear size definitions
are listed here in order of typical decreasing numerical value.
1. The greatest distance between any points on the surface of the grain.
2. The length of the smallest parallelepiped that can be fitted around the grain.
41
3.1 Introduction
(a) Measurement of crystals larger than 0.02 mm
ln (population density)
10
mean size = 0.067 mm
8
6
4
2
0
–2
0
l mm
0.5
1.0
crystal size (length)
1.5
(b) Measurement of crystals larger than 0.10 mm
ln (population density)
10
mean size = 0.205 mm
8
6
4
2
0
l mm
–2
0
0.5
1.0
crystal size (length)
Figure 3.1 Effects of size limits on values of mean crystal size and crystal
size distributions: example of an andesite from Soufriere Hills, Montserrat
(Higgins & Roberge, 2003). The original data have been slightly modified for
this diagram. (a) All crystals larger than 0.02 mm have been counted. Crystals
smaller than this have been observed but not quantified. Hence, the CSD is
undefined below 0.02 mm. The mean size is 0.067 mm. (b) All crystals larger
than 0.10 mm have been measured. The CSD for sizes greater than 0.1 mm is
identical to the upper diagram, but the mean crystal size, 0.205 mm, is almost
three times greater. Hence, if a rock contains crystals of all sizes then the mean
size is meaningless – it just depends on the measurement size limits.
1.5
42
Grain and crystal sizes
3. The mean projected height (mean calliper diameter) of the grain. This can be
visualised at the mean height of the shadow of the grain when it is illuminated
from all different angles. This is commonly used in stereology.
4. The length of the major axis of a best-fit ellipsoid. This ellipsoid has the same
volume and moment of inertia as the grain.
5. The diameter of a sphere of equivalent volume, or more rarely the edge of a cube of
equal volume.
Other size definitions used for unconsolidated sediments are described in
Section 3.3.1. For grains close in shape to a sphere all definitions give similar
sizes. The difference between the measures of size increases with the anisotropy
of the grain. Definitions 2, 3 or 4 are generally similar and should be used. The
definition of size used should always be specified in publications.
3.2 Brief review of theory
A wide range of processes and constraints can affect the crystal or grain size
distribution of rocks, hence the interest in this parameter. For igneous rocks
we must consider the kinetic process of nucleation and initial growth
(Figure 3.2). The driving force for kinetic processes can be expressed as the
undercooling or supersaturation of the system (see Section 3.2.1.2). The resultant crystal populations may be modified by mechanical processes such as
sorting and mixing. Once the kinetic driving force is reduced, the initial or
modified crystal populations will start to equilibrate, although complete equilibrium is never achieved. This process involves the minimisation of the total
energy of the crystal population. The importance of distinguishing kinetic and
Initial
magma
Development of
igneous textures
Nucleation and
growth
Kinetic textures
Sorting and
compaction
Equilibration
Renewed nucleation
and growth
Mechanically-modified
textures
Equilibration
Equilibrium textures
Figure 3.2 Development of igneous textures – role of kinetic, mechanical
and equilibrium processes.
3.2 Brief review of theory
43
equilibrium effects cannot be overstated – it is frequently a source of significant
confusion in the interpretation of CSDs. Melting of igneous rocks will probably
be more controlled by equilibrium textural processes, as it is easier to destroy a
crystal lattice than to make one. But again we must consider mechanical effects
on the crystal population. Kinetics and equilibrium are also important in
metamorphic rocks, but deformation clearly plays a more important role than
in igneous rocks. Finally, I will consider the most generally applicable constraint
on grain and crystal populations, closure: this is the fact that a rock cannot
contain more than 100% crystals or grains. As in geochemistry, this constraint is
commonly acknowledged but it is ignored (Higgins, 2002a).
3.2.1 Primary kinetic crystallisation processes
In many branches of geology the roles of kinetics and equilibrium are important, and that is certainly the case here: kinetics will control the speed at which
crystals can nucleate and grow (or dissolve) and equilibrium will dictate the
final amount of the phases. The most common way of expressing the kinetic
effect is in terms of undercooling: a certain overstepping of equilibrium is
necessary to force the nucleation and/or growth of a new phase. However, the
amount of undercooling encountered during the crystallisation of rocks is
probably strongly dependent on the crystallisation environment and minerals
present, and is not well established so far.
The fundamental functions that control growth textures are simply
expressed: during solidification the nucleation rate is a function of time J (t),
whereas the growth rate is a function of both time and crystal size G (t,l). If
both functions are known, then the evolution of the crystal size distribution
with time, n (t,l) can be determined. Generally speaking we are more interested
in the inverse problem: that is we have measured the final crystal size distribution and we want to know the nucleation and growth rate variations during
solidification. There is no unique solution to this problem, but many attempts
have been made by making assumptions about the form of G (t,l) and J (t)
(e.g. Brandeis & Jaupart, 1987, Marsh, 1988b, Lasaga, 1998, Marsh, 1998,
Zieg & Marsh, 2002).
It is commonly overlooked that significant undercooling is not necessary for
crystallisation: equilibrium requires that the total volume of crystals will
increase during cooling. The texture of the crystals responds to equilibrium
effects to minimise the total energy of the crystal population. This process is
common and has many names: the effect on crystal size is referred to here as
coarsening (see Section 3.2.4). Hence, the final texture of most igneous rocks
will reflect some combination of kinetic and equilibrium effects.
44
Grain and crystal sizes
3.2.1.1 Nucleation of crystals
In a liquid, such as magma or an aqueous solution, atoms are continually being
added or removed from clusters that have a lattice arrangement similar to that
of major mineral phases. Addition of new atoms creates bonds that reduce the
energy of the cluster. However, bonds at the surface of the cluster are unsatisfied and there is a net increase in energy. The sum of these two components
gives the total energy of the cluster. With increasing cluster size the total energy
rises to a peak and then descends (Figure 3.3a; e.g. Dowty, 1980, Markov,
1995). When one of these clusters exceeds a critical radius, r , then it is
energetically stable and a crystal nucleus is formed. Addition of further
atoms to the nucleus is energetically favoured as it reduces the total energy.
This process is termed homogeneous nucleation. The driving force for nucleation is related to the undercooling of the system (see next section). At low
undercooling the energy barrier is not breached, no clusters reach the critical
size and the nucleation rate is zero (Figure 3.3b). With increasing undercooling, the nucleation rate increases rapidly to a peak and then diminishes.
Heterogeneous nucleation occurs on the surface of existing host grains or
bubbles, or on defect structures within crystals. The surface energy of the
cluster is lower, hence both the critical radius and energy barrier are lower
(Figure 3.3a). Hence, formation of stable nuclei occurs at lower undercooling
than for homogeneous nucleation. Mismatches between the crystal lattices of
the new and existing grains will produce elastic stresses that store energy.
Hence, the actual undercooling will generally depend on the degree of crystallographic similarity between the host and the new phase. This is expressed as
the dihedral angle, (see Section 4.2.3; Figure 3.3a).
The necessary undercooling is obviously smallest when the new and host
crystals are mineralogically identical and in the same orientation. However,
in this case the greater ease of nucleation will be balanced by the lower
concentration of the mineral component in the magma surrounding the
growing host crystal.
Crystal nuclei are too small and unstable to be observed – we can record
them only if there has been an interval of growth. Hence, it has been difficult to
study nucleation. Many authors assume that magmas are never completely
liquid and that there is always the possibility of heterogeneous nucleation.
However, as was shown above, the energy difference as expressed by the
dihedral angle may be sufficiently high that homogenous nucleation occurs
even in the presence of solid phases.
New crystals may also form within existing crystals, for instance during
exsolution related to cooling of feldspars. Such crystals can form
45
3.2 Brief review of theory
w
Hi
gh
ed
ium
(a) ΔG
M
Lo
Homogeneous
(θ =180°)
Surface
energy
Critical
radius, r ∗
θ = 90°
cluster
radius
h
Hig
Total
energy
θ = 30°
Me
m
diu
New phase
energy
Lo
w
(b)
nucleation rate normalised
to maximum
θ = 30° θ = 90°
0
Homogeneous
(θ =180°)
undercooling
Figure 3.3 (a) Total energy of an atomic cluster against cluster radius. The
total energy of an atomic cluster is the sum of the surface energy and the new
phase (lattice) energy. For homogeneous nucleation the surface energy is high
and the critical radius is large. For heterogeneous nucleation the surface energy
is lower giving a lower critical radius. Two cases are shown here which are
expressed by dihedral angles of 908 and 308. Although homogeneous and
heterogeneous nucleation processes are commonly presented as opposites, it
is clear that there is a continuum between these two situations. (b) For
homogeneous nucleation at small degrees of undercooling the normalised
nucleation rate is zero. As undercooling increases, it increases rapidly to a
peak and then diminishes. In geological situations most nucleation probably
occurs to the left of the peak. Heterogeneous nucleation reduces the
undercooling necessary for nucleation: a lower energy barrier (DG) leads to
lower dihedral angles and lower undercooling.
discontinuously by nucleation and growth, in the same way as for nucleation
in fluids or on surfaces. However, heterogeneities in crystals may also arise
continuously by a process called spinoidal decomposition (McConnell,
1975). Compositional waves form spontaneously in the mineral lattice: the
46
Grain and crystal sizes
wavelengths are short at low temperatures, but become longer with increasing
temperatures. Such zones may later transform into separate phases which can
be observed as regular exsolution patterns of contrasting phases. The whole
process is not dissimilar to geochemical self-organisation (Ortoleva, 1994).
3.2.1.2 Growth and solution
At its simplest, crystals grow if this results in a reduction in the total energy
of the crystallising system, that is the population of crystals and the material
from which they are crystallising. Similarly, crystals will dissolve also if
this results in a reduction in the total energy. We are principally interested in
crystal growth rates: that is the change in size of the crystal with time.
Unfortunately, although time is the essence of geology, the timing of crystal
growth is generally difficult to measure in the earth. Therefore, experiments
are an important complement to natural studies for determining growth rates.
Crystal growth rates are very important in many industrial processes and
hence there is an immense amount of work in this subject. The problem for
geologists is to determine what aspect of this work is applicable to our field.
Some have seen the range in possibilities, found them wanting and created
their own arbitrary growth laws (e.g. Eberl et al., 1998, Eberl et al., 2002).
Here, I will present a very brief overview of what I consider to be important.
More complete treatment of the subject is available in a number of books and
articles (e.g. Markov, 1995, Marsh, 1998, Jamtveit & Meakin, 1999).
Crystal growth rates depend on the magnitude of the driving force, but the
exact relationship is controlled by the growth mechanism. The driving force
for crystal growth is equal to Dm/kT, where Dm is the chemical potential
difference between the crystallising phase and the host fluid (magma, liquid
or gas), k is the Boltzmann constant and T is the temperature. Dm/kT may be
expressed in many different ways, but the most useful for us is for solutions in
terms of concentrations or enthalpy and temperature (e.g. Lasaga, 1998):
Dm
C
DHsol ðTsat TÞ
¼ ln
¼
kT
C0
kTTsat
where C and C0 are the real and equilibrium concentrations of the solute;
DHsol is the positive enthalpy of solution; Tsat is the saturation temperature
(melting point; mineral liquidus temperature). It is assumed that the system is
ideal and that the enthalpy of the transition is independent of temperature. In
most applications it is easiest to use the undercooling as a measure of driving
force. However, if we want to look at the relationship between crystal size and
growth rate then it is more convenient to use concentration terms.
3.2 Brief review of theory
47
The mechanism of crystal growth is by addition of atoms to the surface. The
rate of addition can be controlled in two ways (e.g. Baronnet, 1984, Lasaga,
1998);
*
*
By the processes of addition of atoms to the surface: interface-controlled growth.
By the transport of the crystal components towards the surface and unwanted
material and latent heat away from the surface: transport-controlled growth. The
actual movement is controlled by diffusion and movement of the fluid with respect
to the crystal (advection).
roughening transition
temperature
The growth shape of a phase depends partly on its physical isotropy or
anisotropy: a physically isotropic material has the same distribution of
atoms in every direction. Isometric or cubic minerals are optically isotropic,
but not physically isotropic. If crystals were physically isotropic then they
would grow in ideal circumstances as spheres: liquids are isotropic and do
grow in this way. However, all crystals are physically anisotropic: hence, the
nature of the crystal–host interface is different in every direction. Of course,
the amount of anisotropy in a crystal is variable and is expressed relative to the
contrast between the crystal and its host. This contrast varies with temperature: below the ‘roughening transition’ temperature crystal faces grow by
spiral growth or two-dimensional nucleation (Figure 3.4). They keep their
orientation and continue to grow as macroscopically flat faces. Above the
roughening transition the energy difference between the crystal and the host
dominates: the crystals grow with rounded surfaces, without facets. In the
temperature
Figure 3.4 Roughening transition (e.g. Baronnet, 1984, Lasaga, 1998,
Bennema et al., 1999). Below this temperature crystals grow with facets;
above it they grow with rounded surfaces. This transition resembles a phase
change. It should be remembered that the roughening transition temperature
may be higher than the saturation temperature (liquidus) of a mineral phase,
in which case the mineral will always crystallise with faces where it does not
impinge on other surfaces.
48
Grain and crystal sizes
geological environment most crystals grow with facets, although these may be
lost where adjacent crystals impinge. Curved faces do occur (e.g. Heyraud &
Metois, 1987), but those produced by solution can generally be distinguished
from those produced during growth by examination of internal growth zones.
Supply of crystal materials depends on several factors: the chemical contrast
between the crystal and its environment and rate of transport of atoms.
Transport includes both diffusion and the relative physical movement of
crystal with respect to the host. The first term is easier to treat: if a crystal
has exactly the same composition as the ‘mother’ material then no diffusive
transport is necessary. An example would be the crystallisation of a pure
substance, such as water. Pure substances are rare in geology: we are normally
concerned with solutions such as magmas.1 Another example is the enlargement of a grain by grain-boundary migration: here, the original and final
materials have the same composition and the transport distance is short.
In more complex situations transport of atoms is controlled by numerous
factors:
*
*
*
*
Physical state of the host material: solid or liquid. Diffusion is much faster in a fluid
phase2 than a solid phase.
Composition of the host material. The overall composition is expressed as the
viscosity. Atoms move more slowly in viscous materials. However, all atoms do
not move at the same rate: generally, small ions move faster than large atoms.
Addition of material to chemically complex minerals requires diffusion of many
different ions, and the overall process is controlled by the slowest atomic species.
Addition of water to magma will reduce the viscosity and increase the diffusion
rates of all species.
The temperature of the system is extremely important: higher temperatures increase
diffusion rates. However, pressure has little effect as solids are almost incompressible.
Relative movement of fluid (advection) from which the crystal is growing with
respect to the crystal. This replenishes the boundary layer that was depleted in the
crystal components by crystallisation and enriched in the rejected components.
3.2.2 Kinetic textures
Crystal populations produced by kinetic processes alone are most commonly
modelled using the Avrami equations (e.g. Lasaga, 1998). However, the
solution of this equation requires knowledge of the variation of the nucleation
rate with time, J (t), the growth rate with time and crystal size, G(t,l). Both of
these are generally unknown and various models of their behaviour have
been proposed. The most common assumption is that the growth rate is
independent of size and hence this variable can be simplified, although Eberl
49
3.2 Brief review of theory
et al. (1998) have proposed that it is proportional to size. It cannot be
overstated that independence of growth rate and size is for kinetic growth
only: it is well established that during coarsening (equilibration) the growth
rate is size dependant (e.g. Voorhees, 1992). Both Gaussian (Lasaga, 1998) and
exponential (Marsh, 1998) models for the variation of growth and nucleation
rates with time have been proposed. One of the most popular population
models uses a steady-state approach and will be discussed first.
3.2.2.1 Steady-state crystallisation models
The nucleation and growth of crystals is a very important industrial process
and much work has been done in this field. One important type of industrial
reaction vessel has a continuous input and output of material, with crystallisation in the vessel. The input material may contain crystals. Crystallisation
occurs in the reactor; hence the output material is richer in crystals than the
input material. The classic study of CSDs in this system by Randolph and
Larson (1971) was adapted to magmatic systems by Marsh (1988b; Figure 3.5).
In both industrial systems and topologically similar open magmatic systems
there is a linear relationship between the natural logarithm of the population
density at size L and that size. For one phase in the system:
(b)
(a) Volcano
e
=
–1
/G
τ
Crystallisation
op
Magma
chamber
Sl
Continuous
flow
ln (population density)
Nucleation density
crystal size
Steady-state crystallisation model
Figure 3.5 (a) Marsh (1988b) proposed that a continuously erupting
volcano approximates a steady-state reactor. Magma with or without
crystals enters the magma chamber from below and partial crystallisation
occurs in the chamber. An equal quantity of magma is withdrawn and erupts.
(b) The CSD of a phase in the magma chamber is a straight line on graph of
ln (population density) against size.
50
Grain and crystal sizes
n0V ðLÞ ¼ n0V ð0ÞeL=Gt
where n0V ðLÞ is the population density of crystals for size L, n0V ð0Þ is the final
nucleation density, G is the growth rate and t is the residence time. The
characteristic length, C, is defined by:
C ¼ Gt
(a) Increasing
nucleation density
ln (population density)
ln (population density)
It is equal to the mean length of all the crystals in a straight CSD that extends
from zero to infinite size. This distribution is linear on a graph of ln(population
density) versus size (L). The intercept is lnðn0V ð0ÞÞ and the slope is 1/C. For
steady-state systems the characteristic length equals the residence time multiplied by the growth rate. There are few geological systems that exactly resemble the original industrial system, but the approach of Marsh (1988b) gave
considerable impetus to the study of CSDs.
This crystallisation model predicts the CSD of a system in a steady state,
which is a single CSD. In geology we are more interested in the dispersion of
parameters in systems. What sort of variation can we then expect in such a
system? If the residence time is constant, but the nucleation density is increasing
then the CSDs will be parallel (Figure 3.6). Increasing nucleation density could
be produced by increasing undercooling. Increasing residence times for constant
nucleation density will make the CSDs swing around a point on the vertical axis.
Such a situation could occur if the magma chamber is expanding. Any combination of these two models is also possible. There is, however, no observational
evidence so far that any geological system actually conforms to these models.
Zieg and Marsh (2002) have taken the linear CSD model described above
and linked it to cooling equations. They used simple cooling models based on
one-dimensional cooling across a single contact and defined nucleation and
growth parameters that relate the mean size of crystals to the cooling interval.
Application of this theory to actual intrusions necessitated the introduction of
an arbitrary growth law, which somewhat weakens the strength of the method.
size
(b) Increasing
residence time
size
Figure 3.6 Theoretical dispersion of CSDs in a continuously fed crystallisation
system with (a) variable nucleation density and (b) residence time.
51
3.2 Brief review of theory
Changes in the crystallisation parameters during solidification, such as cooling rate, may change the characteristics of the CSDs. In general the system will
take time to accommodate the new conditions and during that time the CSD will
be curved on a classic CSD diagram (Marsh, 1988b, Maaloe et al., 1989,
Armienti et al., 1994). Such a transition will change the nucleation rate of the
magma and initially produce a steeper part of the CSD at small sizes. However,
with time the increased nucleation rate will propagate through the CSD as a
step, until finally the CSD is again flat from end to end. Cashman (1993)
proposed that primary curved CSDs could be produced at a constant cooling
rate if there is a change in the nature of the phases precipitated. In this case there
must be crystallisation over a significant period of time in the two-phase field.
3.2.2.2 Batch crystallisation: Marsh (1998) model
Obviously the steady-state approach of Randolf and Larson (1971) cannot be
applied directly to closed-system crystallisation. However, Marsh (1998) also
examined closed systems and discovered that a similar correlation can be
produced under certain circumstances, although for high crystal contents the
CSD is not linear for small crystals (Figure 3.7). In such a system the logarithmic–
linear correlation is produced by exponentially increasing nucleation
density with time, as would be produced if the undercooling of the magma
6
ln (population density)
5
crystal content
4
95%
3
75%
2
50%
25%
1
5%
0
0
0.2
0.4
0.6
crystal size, L /Lm
0.8
1.0
Figure 3.7 Development of closed-system CSDs with increasing crystal
content. Crystal sizes have been normalised to that of the largest crystal.
The turn-down for small sizes reflects the lower magma volume and hence
nucleation rate at high crystal contents. After Marsh (1998).
52
Grain and crystal sizes
was increasing linearly. Just as for the case of the open system, a length can be
derived from the slope of the CSD. However, this ‘characteristic length’ does
not have a clear physical meaning, as was the case for the open system.
Experimental or natural validation of these models is lacking – the only
system studied in detail that might replicate closed-system CSDs, the
Makaopuhi lava lake, had CSDs with constant intercept and variable slopes
(see Section 3.4.1.1; Cashman & Marsh, 1988).
3.2.2.3 Other kinetic population models
Lasaga (1998) has explored the CSDs that are produced by Gaussian-shaped
variations of nucleation and growth rates with time. He found that the growth
rate must peak before the nucleation rate reaches its maximum otherwise
totally unnatural CSDs are produced. He also showed that few CSD
models produce straight CSDs on a classical CSD diagram and that minor
variations in the parameters of the rate variations can produce significant
changes in the CSDs.
Eberl, Kile and co-workers have proposed that kinetic growth rate is
proportional to crystal size (Eberl et al., 1998, Kile et al., 2000, Eberl et al.,
2002). They based this idea on their observation that many CSDs are lognormal. However, in many of their studies it is not clear if the CSDs are truly
lognormal when the data have been stereologically corrected and lower analytical cut-off limits are respected. Approximately lognormal CSDs do exist
(Bindeman, 2003) and are easily produced during equilibration (see
Section 3.2.4) and this is most likely the case for their observations.
Maaloe (1989) has proposed that many CSDs are exponential and that a
growth probability model can explain this distribution. However, this effect
was not enough to explain the curvature of his CSDs and hence he also
invoked a two-stage model of crystallisation.
3.2.3 Mechanically modified textures
Populations of crystals can be modified by mechanical processes, such as
compaction, sorting and mixing. Some of these processes are closed, in that
the total mass of the system is unchanged, others are open to the addition of
crystals and magma.
3.2.3.1 Compaction and filter pressing
A common process that can change CSDs is compaction (e.g. McBirney &
Hunter, 1995, Hunter, 1996, Jerram et al., 1996, Higgins, 1998). This process
starts as mechanical compaction, in which reorganisation of the grains enables
3.2 Brief review of theory
53
ln (population density)
Pressure-solution
compaction
Simple
compaction
Or
igi
na
lC
SD
size
Figure 3.8 Simple compaction changes all the population density values
evenly. Pressure-solution compaction will probably result in the preferential
resorption of small crystals and will hence have similarities with coarsening
processes (Higgins, 1998).
pore fluid to be squeezed out. This will change the population density of each
point on the CSD, pushing it upwards evenly (Figure 3.8). It will result in an
increase in the intercept, but not the slope of the CSD. For instance, 50%
compaction will increase the intercept by 0.7 (¼ ln(1/0.5)). All minerals that
were present during compaction should show the same displacement of their
CSDs. It is possible that the finest grains may be removed with the fluid – this
will produce a lower population density at the left end of the CSD. Mechanical
compaction is limited to a maximum of about 50% by the original coherence
of the magma and does not lead to the development of a significant foliation
(Higgins, 1991). Filter pressing is the same process except that we are more
concerned with the expulsed liquid.
Higher degrees of compaction must involve changes in the shape of the
grains. Hunter (1996) has reviewed the different mechanisms by which crystals
can change shape during compaction. For cumulate rocks with a small amount
of melt, melt-enhanced diffusive creep will be the most important mechanism.
Material will be dissolved from surfaces with higher energies, such as those in
contact with other grains. This is commonly termed the pressure-solution
effect (Figure 3.9). Meurer and Boudreau (1998) proposed a compaction
model where mechanical rotation of grains is accompanied by pressure
solution of the ends of grains not orthogonal to the direction of maximum
54
Grain and crystal sizes
(a)
(b)
(c)
Figure 3.9 Textural changes during pressure-solution (textural equilibrium)
compaction. After Hunter (1996). (a) Initial impingement texture. (b) Early
pressure solution. (c) Porosity eliminated by pressure solution.
stress. This process could lead to much larger degrees of compaction and the
development of a significant foliation. Although this model has not been
quantified, it has many aspects in common with coarsening – parts of grains
go into solution because they have higher surface energy. Higgins (1998)
proposed that the CSDs produced by pressure-solution compaction are
similar to those produced by coarsening (Figure 3.8). However, the two
processes are not exactly equal as crystals of all sizes can be dissolved by
pressure solution.
Pressure solution is driven by minimisation of the total energy of the system
as is the coarsening process. Hence, if CSDs are modified in the same way then
there should be a correlation between the characteristic length of the CSD and
the quality (degree) of foliation. Higgins (1998) found that a few samples
showed this effect, but most did not. Hence the process may only be locally
important.
There is another effect by which this process may be recognised: during
compaction-pressure-solution the ends of crystals not orthogonal to the direction of compaction will be dissolved and hence the shape anisotropy will no
longer correspond to the lattice orientation.
Although compaction pressure solution may be able to produce a significant
foliation, flow of crystal-rich magma is a much more efficient process (e.g.
Higgins, 1991, Ildefonse et al., 1992, Nicolas, 1992, Nicolas & Ildefonse, 1996).
The two processes can be distinguished if the magma contains prismatic
crystals as flow produces a lineation. In many cumulate rocks plagioclase is
measured and this is generally tabular, hence without a lineation. Simple
magmatic flow does not change the CSDs, and hence its action can only be
revealed by textural measures of crystal orientation. Nicolas and Ildefonse
(1996) have proposed that pressure-solution plays a significant role during
flow of magma by removing obstacles and entanglements. In this case it would
affect CSDs in the same way as compaction-pressure solution.
3.2 Brief review of theory
55
3.2.3.2 Crystal accumulation
Crystal accumulation is the process of the separation of crystals from magma.
There are many different mechanisms that can achieve this: filter pressing (gas
exsolution), flowage differentiation (Bagnold effect), gravity separation and
crystallisation on conduit walls. Although he did not model this process,
Marsh (1988b) suggested that accumulation of crystals by gravity will produce
curved CSDs on a classical CSD diagram. Higgins (2002b) has developed a
simple model for crystal accumulation based on Stokes’ law (Figure 3.10).
For the purposes of modelling, this process is considered here only in its
simplest form – the separation of crystals and liquid in response to differences
in density. Stokes’ law is used to calculate the CSDs in a zone of accumulation
of crystals. Many magmas are not Newtonian, especially at high crystal
concentrations, but their yield strength under different conditions is not
well known (Naslund & McBirney, 1996). In addition, Stokes’ law is for
spheres. Hence, Stokes’ law is not totally applicable in a magma chamber
and this model should be viewed as a point of departure, which is probably
more accurate for the initial stages of the process when crystal concentrations
are low.
Column of magma
100 cm
Displacement of
crystals follows
Stokes’ law
1 cm
Accumulation
zone
Figure 3.10 Simple modelling of accumulation of crystals using Stokes’ law
(Higgins, 2002b). The position of each crystal after a discrete time interval is
calculated from Stokes’ law. The crystals are allowed to accumulate in a zone
at the bottom of the column. Flotation of crystals could be modelled in the
same way.
56
Grain and crystal sizes
A population of crystals is placed in a column of magma and allowed to
settle under gravity. Their vertical positions at subsequent times are calculated
using Stokes’ law:
vt ¼
2gDrr2
9
where vt ¼ terminal velocity, g ¼ gravitational constant, Dr ¼ density difference between the magma and the crystals, r ¼ crystal radius and ¼ viscosity.
The CSDs of those crystals present at the bottom of the column is that of the
cumulate rock. For simplicity, a fixed accumulation zone is defined that can
hold all the crystals in the column. A low initial concentration of the crystals
ensures that the collection zone never fills up and that the small number of
crystals still in the magma does not have a large effect on the cumulate CSD.
This model was applied by Higgins (2002b) to the settling of plagioclase and
olivine in a basaltic magma (Figure 3.11). The model begins with a population
of spherical olivine and plagioclase crystals that has a straight-line CSD with
slopes and intercepts similar to those observed in the Kiglapait intrusion
(Higgins, 2002b), but at only 1% of the crystal concentration. These crystals
are initially randomly distributed in a one-metre-high column. Initially the
CSDs remain linear and rotate about the intercept. After one hour 35% of the
olivine but only 4% of the plagioclase has been precipitated. Later, the right
ends of the CSDs reach a limit as all the crystals have been precipitated. The
linear section then propagates towards the origin. At later stages the CSD has
a pronounced turn-down for small crystals, but eventually becomes a straight
line parallel to the original CSD. At the end the whole process is equivalent to
simple compaction. In a real situation, for example at the base of a magma
chamber, crystal settling is never allowed to go to completion: new magma will
sweep in displacing the older, fractioned magma. Hence, the early part of this
model is probably the most applicable.
3.2.3.3 Mixing of magmas and crystal populations
Mixing of magmas is a very common process – many plutonic and volcanic
rocks show abundant evidence for this process. The term mixing is usually
reserved for the extreme limit of the process that produces a homogeneous
product. Mingling refers to a less extreme process where the two components
can still be distinguished. However, this is just a question of scale. CSDs are a
bulk measure and hence will sample macroscopic regions that may not be
completely homogeneous.
Higgins (1996b) showed that addition of two straight CSDs with contrasting
slopes and intercepts will give a CSD with a steep slope at small sizes and a
57
3.2 Brief review of theory
Infinite time
0
(b)
Olivine
35%
0.1 h
4
36%
1h –
– 3%
ial
2
10 h –
–5
Init
–10
0
Plagioclase
100 h
1h –
–5
Infinite time
0
100 h
10 h
ial
–10
6
8
size (mm)
10
12
4%
Init
0
2
4
6
8
size (mm)
0.1
h
10
12
(d)
(c)
Top
schematic layer
Original CSDs
Ol
Pla
gio
ivi
ne
cla
se
Bottom
Olivine
Plagio
olivine
plagioclase
clase
Original CSDs
volumetric phase abundance
ln (population density)
(a)
Top Pla
g
Bottom
ioc
las
e
e
in
liv
O
Top
Bottom
characteristic length
Figure 3.11 Modelling of accumulation of plagioclase and olivine in a
basaltic magma using Stokes’ law (Higgins, 2002b). The following
parameters were used: viscosity ¼ 400 Pas, magma density ¼ 2620 kg m3,
olivine density ¼ 3500 kg m3, plagioclase density ¼ 2700 kg m3. The
magma column was one metre high. The CSDs are for the bottom 1 cm.
(a, b) The evolution of plagioclase and olivine CSDs from an initial, straight
CSD. (c) Schematic effect of settling of plagioclase and olivine in the column.
(d) Effect of settling on the amount of a phase (volumetric phase abundance)
and the characteristic length (1/slope) of the CSD.
more gentle slope at larger sizes. Summation of the two linear CSDs gives a
curved CSD with two linear segments because of the logarithmic scale of the
population density axis (Figure 3.12). It is clear that the slope of the original
CSD of both magmas can be recovered from the linear segments even when
one component is only 1% of the total. However, it should be noted that
although mixing of two populations of crystals will not change the slope of
their CSDs, the associated intercepts will be lower than that of the original
magmas. Therefore, if the Marsh (1998) model is accepted then residence times
can be determined for each population, but no meaning can be attached to the
value of the intercept, unless it can be corrected for the effects of dilution by the
58
Grain and crystal sizes
0% %
% 1
1
50
ln (population density) (mm–4)
10
5
Pure A
0%
0
50%
90%
–5
99%
10
0%
Pure B
Percentage of B
–10
0
0.2
0.4
0.6
0.8
1.0
size (mm)
Figure 3.12 Addition of two linear CSDs (after Higgins, 1996b). The
original slopes of the two components are preserved for large and small
crystals, but not their intercepts.
other magma. Such corrections can be determined from the mixing proportions or the initial crystallinity of the magma. Further growth of crystals
following mixing will generate a third population of crystals, represented by
another linear segment of the CSD at the left of the diagram. The existing
crystals will grow rims and their CSDs will be displaced towards the right of
the diagram, but the slopes of the linear segments will be preserved. It should
be noted that the deconvolution method described above can also be applied to
multiple populations of crystals that grew sequentially from the same magma,
if such processes produce straight CSDs.
If two magmas are mixed together then the volumetric proportion of a phase
Vtotal in the final product can be determined from (Higgins, 1996b):
Vtotal ¼ VA XA þ VB XB
where VA, VB are the volumetric abundances in end-member magmas A and B
and XA, XB are the volumetric proportions of the magmas. This equation can
be recast as:
XA ¼
Vtotal VB
VA VB
Therefore, if the initial volumetric phase abundance of both magmas can be
estimated and the final volumetric phase abundance of the mixture is known,
then the proportions of the two magmas can be calculated. Unfortunately, the
59
3.2 Brief review of theory
initial volumetric phase abundance of both magmas is not determined easily.
The apparent volumetric phase abundance of the two magmas can be calculated from the modelled slope and intercept of the CSD for each crystal
population and the crystal shape. However, this is not the initial value before
mixing, but a lower value produced by ‘diluting’ the initial volumetric phase
abundance with the other magma.
3.2.4 Textural effects of equilibration
At the surface of a crystal there are unsatisfied bonds which have a greater
energy than those within the crystal: this excess energy is minimised as the rock
texture equilibrates. Although complete equilibrium is never reached the
texture will tend towards a population of ever larger, even-sized crystals.
Crystal populations produced by kinetic processes, even if modified by
mechanical processes, are far from equilibrium as they contain crystals with
a wide range of sizes: smaller crystals will have a higher surface area and hence
energy per unit volume than larger crystals (Figure 3.13). Equilibration will
reduce the total energy, which in a defect-free crystal is related to the total
surface area of the crystals. The process generally starts at grain boundaries
where the dihedral angles will adjust to uniform values (see Section 4.2.3). The
process will continue into the rest of the crystal boundaries, giving them an
8
1
2
Number of crystals
Volume of crystals
Surface area of crystals
1
1
1
Higher
Equilibrium
concentration
Lower
Figure 3.13 Spherical crystals of the same phase of two different sizes. Both
populations have the same total volume but different crystal numbers and
surface areas. Equilibration will tend to reduce the number of small crystals
and assemble their material into larger crystals.
60
Grain and crystal sizes
even curvature. Finally, the size population of the crystals will be affected by
the solution of small crystals and the growth of larger crystals, which is
discussed here. There are aspects of crystals other then size that will also
contribute to excess energy: complex crystal boundary shapes and lattice
defects are probably the most important. Minimisation of the energy associated with these structures may also power coarsening processes, although
initially the process may take a different form (see Section 4.2.2).
The process of textural equilibration is commonly observed and carries
many names, some with slightly different definitions: coarsening (Higgins,
1998), textural maturation, textural equilibrium (Elliott & Cheadle, 1997),
Ostwald ripening (e.g. Voorhees, 1992) crystal aging (Boudreau, 1995), competitive particle growth (Ortoleva, 1994), annealing and fines destruction (Marsh,
1988b). Metamorphic petrologists also use a wide range of terms for this
general process: recrystallisation (a somewhat ambiguous term), grain-boundary
area reduction and grain-boundary migration. The term coarsening will be used
here as it applied to CSDs.
3.2.4.1 Energy considerations
Coarsening is an important process in metallurgy, ceramics and other industrial fields (see the review by Voorhees, 1992). Most authors consider that it is
important in metamorphic rocks (Nemchin et al., 2001), although there are
some dissenters (e.g. Carlson, 1999). The role of coarsening in the development
of igneous rocks is much less well established but seems to be important in
plutonic and some volcanic systems (e.g. Hunter, 1996, McBirney & Nicolas,
1997, Higgins, 1998).
The total surface energy of a grain is related to its area by the interfacial free
energy, g. The equilibrium concentration of an essential element for a crystal of
radius r, Ce ðrÞ, in the material surrounding the crystal will have to reflect these
differences in energy: smaller crystals will have higher values than larger
crystals. This can be calculated with the Gibbs–Kelvin equation for isotropic
materials (e.g. Marqusee & Ross, 1983):
Ce ðrÞ ¼ Ce ð1Þea=r
where
a¼
2gw
kT
and w is the unit cell volume of the mineral, k is the Boltzmann constant and
T is the temperature in K. The value of Ce ðrÞ decreases slightly at higher
temperatures. However, the most important control, apart from radius, is
3.2 Brief review of theory
(a)
Olivine
γ high
61
(b)
plag
coarsened texture
original texture
(c)
Magma
γ low
(d)
plag
Figure 3.14 The textural expression of differences in interfacial energy, g,
between two phases during equilibration. (a) If plagioclase crystals are
enclosed by an olivine oikocryst, then the interfacial energy difference will
be large. (b) With equilibration the plagioclase chadocrysts will become
rounded and coarsened. (c) If plagioclase crystals are immersed in a basaltic
magma then the interfacial energy difference is smaller, as the atomic
arrangements are less contrasting. (d) With equilibration the plagioclase
crystals will be less coarsened and rounded.
the interfacial energy, g. This parameter can have a significant effect on rock
textures (Figure 3.14).
If crystals of different sizes are adjacent then the equilibrium concentrations
of essential elements near the surfaces of the crystals will be different. Clearly
this concentration gradient cannot be maintained and diffusion will tend to
even out the concentrations. Once the concentration of an element is changed
the crystal will either grow or dissolve. Normally, the smaller crystals will
dissolve and the large crystal will grow. All the parameters of this equation are
well known except the interfacial energy. This depends on the nature of the
surrounding material. Relative values can be derived from the dihedral angles
of crystals (see Section 4.2.3), but absolute values are much less well known:
Nemchin et al. (2001) reported a range of a factor of 80 for the interfacial
energy of zircon.
For igneous systems there is another way of looking at the problem.
Variation of the rates of nucleation and growth with undercooling are well
known (Figure 3.15). Coarsening can only occur where the nucleation rate is
62
Grain and crystal sizes
normalised nucleation
or growth rate
Coarsening
region
Nucleation rate
Growth rate
normalised undercooling
Figure 3.15 In igneous rocks coarsening occurs at low undercooling, close
to the liquidus, where the nucleation rate is zero, but the growth rate is
significant.
zero, but the growth rate is significant. This is close to the liquidus of that
mineral. This can also be visualised in terms of the critical radius of atomic
clusters: close to the liquidus even very large, macroscopic clusters are not
stable.
Material is transferred from grains smaller than a critical radius to the larger
grains by several different diffusive mechanisms (Figure 3.16; see review in
Hunter, 1996). If there is no liquid present then diffusion may be within the
body of the grains or, more likely, along grain boundaries, both of which are
slow. If a liquid is present then diffusion in the liquid dominates any withingrain diffusion. The latter is the most important mechanism for igneous rocks.
The term grain-boundary migration is frequently used in igneous and metamorphic petrology: it refers to diffusion either along grain boundaries or over
a short distance via an intergranular liquid.
3.2.4.2 Lifshitz–Slyozov–Wagner coarsening model
Although the overall thermodynamic driving forces of coarsening are well
known and understood the kinetics of the process are complex and the subject
of much research, especially in the field of materials science (Voorhees, 1992).
Unfortunately, the number of variables is very large and most of the systems
63
3.2 Brief review of theory
Along grain boundaries
Within grains
Via a fluid / liquid
Figure 3.16 Textural changes associated with diffusive creep involving
different pathways for a system under a generalised vertical stress.
Diffusion can be along grain boundaries (Coble creep); within the
grains (Nabarro–Herring creep); or via an intergranular fluid/liquid. After
Hunter (1996).
investigated are not appropriate to geological conditions. Nevertheless, a good
starting point is the earliest and most well-known solution, that formulated by
Lifshitz, Slyozov (1961) and others, known as the Lifshitz–Slyozov–Wagner
(LSW) theory. The solution is for dilute systems, where there is assumed to be
no interaction between crystals: all communicate directly and instantaneously
with a uniform fluid. The crystals are assumed to be spherical. Most applications of LSW theory (and modifications of the theory) have been concerned
with the relationship between mean crystal size and time. However, in plutonic
rocks solidification time is not well constrained. Under these circumstances an
equation relating growth rate to crystal size is more pertinent (Lifshitz &
Slyozov, 1961).
dr
k 1
1
¼
dt
r r
r
where dr/dt is the growth rate, r is the critical radius, r is the crystal radius
and k is a rate constant.
A graph of growth rates against radius shows that crystals below the critical
radius are strongly resorbed and their material transferred to crystals slightly
larger than the critical radius (Figure 3.17). The growth rate of large crystals
tends to zero. The rate constant k is dependent on temperature and its value is
not known.
Simple modelling of the effects of LSW coarsening on an initially straight
CSD are shown in Figure 3.18a. There is a progressive loss of very small
64
Grain and crystal sizes
1.0
CN model
Growth
growth or solution rate
maximum growth rate
LSW Model
0
Critical radius
–1.0
Solution
–2.0
0
5
10
radius / critical radius
15
20
Figure 3.17 Growth rate variations for the LSW model and the Communicating
Neighbours (CN) model (Higgins, 1998).
(b) Textural coarsening
CN model
ln (population density)
ln (population density)
(a) Textural coarsening
LSW model
size
Open system
Closed system
size
Figure 3.18 Simple stochastic modelling of the development of CSDs
during LSW and CN coarsening (Higgins 1998). (a) For LSW coarsening
the critical radius increases with each time step, but the rate constant is
unchanged. (b) For CN modelling both the critical radius and the rate
constant increase with each time step. Higgins (1998) proposed that opensystem coarsening could reduce the slope of the CSD yet further.
crystals, but the actual numbers of larger crystals do not change very much. It
is frequently emphasised that LSW Ostwald ripening will ultimately produce a
CSD normalised to mean grain size that has a shape which is only dependent
on the rate-controlling growth mechanism (Lifshitz & Slyozov, 1961).
3.2 Brief review of theory
65
However, this is probably not very useful in geological situations as complete
equilibration needs constant conditions for a long time and is hence rarely
achieved. Another problem is that the quality of much CSD data, especially
for small sizes, is not sufficient to distinguish the different mechanisms.
3.2.4.3 Communicating Neighbours coarsening model
The limitations of the LSW equation are well known; hence it is not surprising
that it is unable to model the observed CSDs. Numerous modifications have
been proposed, such as extensions to high volume fractions (see review by
Voorhees, 1992). However, as DeHoff (1991) has pointed out
‘The literature abounds with theoretical treatments of coarsening for an additional
important reason: the existing theories do not work.’
DeHoff proposed a new theory that he called ‘Communicating Neighbours’
(CN), that was applied to plutonic rocks by Higgins (1998).
CN theory is based on the observation that in many laboratory studies the
growth rate of each crystal is not solely controlled by its size, but appears to be
characteristic of each crystal, an effect called growth-rate dispersion. This
suggests that the position of each crystal is important during growth and
hence that crystals communicate with each other and not only with some
uniform fluid. Of course, the mechanism for communication as in the LSW
theory is by diffusion. DeHoff (1991) has formulated a geometrically general
theory using this principle. The most readily applicable equation relates
growth rate and crystal radius.
dr
1
1 1
¼k
dt
l r r
where dr/dt is the growth rate, k is a rate constant (different from that in the
LSW theory), h1=li is the harmonic mean of the intercrystal distance l, r is the
critical radius and r is the crystal radius. This equation illustrates the fundamental difference between the LSW and CN theories: in the former the diffusion length scale is a property of the particle itself, hence the 1/r dependence,
whereas in the CN theory it is dependent on the distances to the neighbouring
particles. This equation is for crystal growth controlled by diffusion. A similar
equation also applies to interface-controlled growth (DeHoff, 1984).
A graph of growth rate against radius for the CN equation shows that
crystals below the critical radius are resorbed rapidly, although less so than
with the LSW equation. The big difference is for crystals larger than the critical
radius, where the growth rate increases asymptotically with radius
66
Grain and crystal sizes
(Figure 3.17). This equation is much more interesting than the LSW equation
as the growth rate is constant for crystals much larger than the critical radius,
and will then depend on undercooling, amongst other factors.
Higgins (1998) also applied stochastic modelling to the CN equation
(Figure 3.18). The critical radius and rate constant were increased with equilibration to achieve the best fit to CSD data from the Kiglapait intrusion. The
match for the right side of the CSD diagram is much better than the LSW
model, although the turn-downs at small lengths are not as steep as observed.
It should be noted that the common practice of plotting linear crystal population densities against length conceals the variations in the number of large
crystals (e.g. Cashman & Ferry, 1988). Hence, the difference between the CN
and LSW theories is not so clear on these diagrams.
3.2.4.4 Open systems and fluid focusing
The simplest models of coarsening assume that the system is closed. That is, all
material that is dissolved from the small crystals will be precipitated on the
larger crystals and hence the volumetric proportion of the phase is constant.
However, the transfer of material is via a liquid phase and this phase can migrate
into or out of a rock. Hence, the process of coarsening may be an open process.
Open-system processes may lead to local increases in the volumetric abundance of a phase. For example, in a crystal mush coarsening enhances the
permeability by removing small crystals that may block channels. Increased
permeability will, in turn, lead to focusing of fluid flow, further growth of
megacrysts along the channels and positive feedback (Higgins, 1999). The
channels formed by these interactions will appear as linear regions and nests
of crystals in the two-dimensional surface of the outcrop (Figure 3.19).
3.2.4.5 Adcumulus growth
Many authors consider adcumulus growth to be an important process in the
solidification of plutonic rocks, but the exact process has never been clearly
defined. It is generally assumed that all crystals grow at the same rate
(Figure 3.20). Hunter (1996) discouraged the use of this term because he felt
that it was a model dependent term and not a descriptive term. He considered
that adcumulus growth could not be distinguished from compaction. I propose
that there is also a large component of coarsening involved, perhaps combined in
some rocks by pressure solution compaction (Higgins, 2002b). Removal of small
crystals ensures that permeability is maintained even at low degrees of porosity,
hence allowing the development of rocks with little or no trapped liquid.
3.2 Brief review of theory
Coarsened
grain
Coarsened
grain
Figure 3.19 Open-system coarsening in crystal mushes. Coarsening removes
small crystals and focuses flow. Very coarsened grains (megacrysts) form along
the channels.
(a) Adcumulus growth
(b) Textural coarsening
Final crystal
Initial crystal
Overgrowth
Figure 3.20 (a) Although the exact process of adcumulus growth has not
been specified it is usually assumed that all crystals grow equally. (b) During
coarsening small crystals dissolve and large crystals grow simultaneously.
Permeability is maintained even at low porosity as small crystals that tend to
block circulation are removed.
67
68
Grain and crystal sizes
3.2.4.6 Coarsening in almost completely solid materials: metamorphic
and sub-liquidus plutonic rocks
Coarsening occurs in metamorphic rocks where it is commonly referred to as
‘grain-boundary area reduction’ (GBAR) or Ostwald ripening. In deformed
metamorphic rocks the major difference from igneous rocks is that excess energy
is derived from unsatisfied bonds both at the surface of the crystals and around
defects within the crystals. The balance of these factors is controlled by the grain
size distribution, grain shape and defect concentration. All will be reduced as
the texture equilibrates (see Sections 3.2.5 and 4.2.2). Coarsening can occur
during deformation, but its effect is counteracted by grain size reduction
processes. GBAR is most evident in rocks that have undergone ‘static recrystallisation’, generally after deformation has ceased. If the process is to produce
measurable effects then a fluid is necessary: one dominated by water at lower
temperatures and by silicate at higher temperatures.
Coarsening of crystals suspended in a liquid, or free to move, is clearly very
different from coarsening in a material that is almost completely solid: the
presence of an independent second phase (or phases) may ‘pin’ the position of
the grain boundary and limit coarsening in that direction (Berger, 2004, Berger &
Herwegh, 2004). Such a situation occurs in all polymineralic rocks and may be
important even in almost monomineralic rocks that only contain a small
amount of a second phase. This effect is sometimes referred to as Zenerinfluenced coarsening. Modelling of the influence of second phases suggest
that such CSDs have a lognormal distribution, rather than the Gaussian
distribution produced by the LSW models.
The physical processes governing GBAR are identical to those of coarsening
described previously. However, the saturation temperature of a phase in a
metamorphic rock does not correspond to its liquidus temperature in magma,
but to its upper stability limit. Hence, coarsening should be most important
at higher temperatures, close to the stability limit of the phase. In metamorphic
rocks kinetic effects must always be considered, particularly where fluid is not
abundant. Hence, there has been much discussion concerning the applicability
of coarsening and GBAR (Miyazaki, 1996, Carlson, 1999, Miyazaki, 2000).
3.2.5 Crystal deformation and fragmentation
3.2.5.1 Crystal deformation
Solid-state mechanical deformation of crystals and subsequent recovery of
strain energy can result in grain size reduction (e.g. Karato & Wenk, 2002).
This process differs from cataclasis in that at no point are the crystals
69
3.2 Brief review of theory
mechanically broken. Crystals deform in the solid state by the production and
migration of lattice defects.
Lattice dislocations are places in a crystal lattice where there is a mismatch
between lattice planes (Figure 3.21). In an edge dislocation one lattice plane
terminates and the space is adjusted by irregularities in the lattice. The atoms
at the end of the lattice plane have unsatisfied bonds. In a screw dislocation the
mismatch is restricted to a single point and the lattice is rotated helically about
this point. Another type of lattice defect is a vacancy: here an atom is missing
and the surrounding chemical bonds are unsatisfied. In both cases there is
excess energy associated with the lattice defect.
Lattice dislocations have an unsatisfied bond that can be passed along to
adjacent atoms readily in response to stress of the lattice (Figure 3.22). The
(a) Dislocation
(b) Vacancy
Figure 3.21 Lattice defects in crystals. (a) Dislocations are places within a
crystal where a lattice plane terminates. The mismatch is accommodated by an
irregular unit cell or cells. (b) Vacancies are lattice positions that are unoccupied.
(a) Movement mechanism
of lattice defects
(b) Initial
lattice
(c) Defect is
generated
(d) Defect
migrates
(e) Defect
migrates
(f) Final
lattice
Figure 3.22 (a) Movement of dislocations in a crystal lattice. (b) The initial
lattice has a step on the left. (c–e) A lattice dislocation can move across the
lattice propagating a change in the shape of the crystal. (f) The final lattice has
a step on the right.
70
Grain and crystal sizes
overall effect of the passage of a dislocation through a lattice is that the
crystal will change shape. Clearly edge dislocations are much more efficient
than screw dislocations as a whole part of the lattice can be displaced.
Dislocations can move along a number of different lattice planes. If several
dislocations become entangled then movement may cease. This is the cause of
strain hardening in materials. Such tangles can be removed if vacancies are
allowed to jump over the entanglement. This effect is enhanced at higher
temperatures.
The unfulfilled chemical bonds associated with lattice defects have a higher
energy than other bonds in the lattice. Hence, the energy of a crystal can be
minimised by a reduction in the number density of the defects. This is termed
recovery. If the defects migrate to the edge of the crystal then the grain size is
unchanged. However, it is generally more energetically favourable to accumulate the defects along surfaces within the crystals and hence form new, smaller
subgrains. Each subgrain will have a slightly different orientation from its
neighbours. The overall result of deformation and recovery is the subdivision
of larger grains into a mosaic of smaller grains. There is some evidence that the
CSDs of grain size reduced rocks may be fractal, that is they have a power-law
distribution (e.g. Armienti & Tarquini, 2002).
A much more limited amount of deformation can be accommodated by
twinning (Karato & Wenk, 2002). This effect is commonly seen in plagioclase
and calcite, but can occur in a number of other minerals. Crystals can also
deform by migration of lattice vacancies. This process is more important at
higher temperatures.
3.2.5.2 Crystal fragmentation (cataclasis)
Fragmentation or cataclasis is the mechanical breakage of crystals in response
to strain. It occurs if a crystal cannot deform fast enough by the production
and migration of lattice defects. Hence, it occurs in rocks that are strained
rapidly, or at lower temperatures. The exact threshold values for rapid strain
are variable: some minerals, such as calcite are very weak and plastically
deform easily. Other minerals, such as diamond, are strong, but may be broken
if the emplacement process is sufficiently violent.
Strain can have an external or internal origin. External strain results from
the deformation of the matrix, most commonly by other solid phases that are
in contact with the grain, but it is also possible that rapid deformation of a very
viscous liquid could induce fragmentation. Internal strain comes from changes
in the volume of different parts of the crystal in response to changes in pressure
or temperature. The most common source of internal strain is the effect of
pressure on large fluid inclusions. The liquid and/or gas in inclusions has a very
3.2 Brief review of theory
Figure 3.23
71
A fractal model for fragmentation after Turcotte (1992).
different compressibility to that of the crystal and hence pressure changes can
generate large internal strains (Allen & McPhie, 2003).
One of the earliest models of rock fragmentation was developed by
Kolmogorov (1941) and Epstein (1947). They proposed that breakage is a
series of discrete events, and that during these events the probability of breakage of any clast is constant and independent of size. This model produces a
lognormal size distribution by length. However, industrial crushing and grinding commonly produces fragments with the Rosin–Rammler distribution
(Kotov & Berendsen, 2002). This results when fracture is controlled by initial
flaws that have a Poisson distribution. This distribution is difficult to test and
hence is transformed to the Weibull distribution (Kotov & Berendsen, 2002).
Finally, the most recent fragmentation model is the fractal distribution.
The mathematical concept of fractional dimensions – fractals – is recent
(Mandelbrot, 1982), but has had a remarkable impact on the natural sciences,
particularly in geology (Turcotte, 1992). Indeed, it is now commonly assumed
that most geological structures, including rock textures, are fundamentally
fractal. A simple model of fracturing can give a fractal size distribution
(Figure 3.23). Two diagonally opposite cubes are retained at each scale, and
the other blocks divided into blocks with half the linear dimension. The fractal
size dimension is 2.58 (Turcotte, 1992).
3.2.6 Closure in grain size distributions
The crystal content of a rock cannot exceed 100%; hence, as with chemical
analyses, we must always be aware of the closure problem. Here the problem is
treated for crystal size distributions, but a similar approach can be used for
72
Grain and crystal sizes
grain size distributions. Closure for an individual phase can also occur at less
than 100% crystals. For instance, if a rock is made of 50% plagioclase and
50% olivine then closed-system processes, such as closed coarsening, cannot
change the phase proportions – they are each fixed at a maximum of 50%. This
effect must be considered in all CSD studies, both of volcanic and plutonic
rocks (Higgins, 2002a).
We are concerned here with the effects of constant phase proportion on
CSDs; however, Higgins (2002a) showed that even quite large variations in
volumetric phase proportions can show the same effects. Clearly, closure is just
a special case of this more general problem, where the volumetric phase
proportion is equal to one.
For a population of crystals with constant shape, the volumetric proportion
of phase i, Vi , can be calculated by integration of the volume of all the crystals,
following Higgins (2002a):
Vi ¼ s
Z1
n0Vi ðLÞL3 dL
0
where s ¼ shape factor of phase i, n0Vi ðLÞ ¼ population density of crystals of
phase i for size L. The shape factor is equal to the ratio of the actual volume of
the grain divided by the volume of a cube that encloses the grain, L3. It can be
expanded into a more applicable form as follows
s ¼ ð1 Oð1 p=6ÞÞIS=L2
where O is the roundness factor, which varies from 0 for rectangular parallelepipeds to 1 for a triaxial ellipsoid. S, I and L are the short, intermediate and
long dimensions of the parallelepiped or ellipsoid. Obviously most crystals
have a more complex shape, but we are concerned here just with a statistical
measure. However, the above equation should not be applied to crystals with
concave surfaces or holes, as they do not necessarily have volumes that lie
between that of an ellipsoid and an equivalent parallelepiped.
For straight CSDs (semi-logarithmic) with intercept n0Vi ð0Þ and characteristic length C (C ¼ 1/slope) a simplified equation can be used
Vi ¼ 6sn0Vi ð0ÞC4i
It should be emphasised that volumetric phase proportions, Vi , calculated
using these equations are sensitive to errors in the shape factor – it is common
to find errors up to a factor of two. Hence the actual phase volume should be
determined from the total area of the crystals (see Section 2.7).
73
3.2 Brief review of theory
The correlation between the slope and intercept of straight CSDs has been
noted by many authors (e.g. Higgins, 1999, Higgins, 2002a, Zieg & Marsh,
2002). Zieg and Marsh (2002) linked slope and intercept by a ‘modal normalisation factor’ which incorporates the inverse of the modal abundance and a
shape factor. They attribute this relationship to scaling laws, but it is clear that
it is just a geometric relationship referred to here as closure.
At low volumetric phase proportion, for example 1%, the slope of a straight
CSD can change independently of the intercept (Figure 3.24a). For example, if
all crystals grow at the same rate and nucleation increases exponentially, then
the CSD will move up without changing slope. Once the crystal content
reaches 100% then closure is reached and the CSD is locked. It can only
move if the ratio between crystals of different sizes is changed. That is, some
crystals must become smaller if others are to enlarge.
A family of straight CSDs for 100% crystallised material defines a fan
(Figure 3.24b; Higgins, 2002a, Zieg & Marsh, 2002). Portions of this fan
appear to describe rotation of the CSD around a point commonly close to
the left of the diagram. The CSDs together outline a concave-up envelope.
0
10
5
10 15
size (mm)
20
25
characteristic length (mm)
–15
0
5
10 15
size (mm)
20
25
2
1:
e
0
4
er
1%
–15
–10
6
h
:4 4:4 Sp
1:2 1:
0%
10 %
–10
–5
8
4
–5
0
1:1:
ln (population density) (mm–4)
5
10
ln (population density) (mm–4)
(c) 100% closure limits for
straight CSDs with
different crystal shapes
(b) Family of straight CSDs
with 100% cubic crystals
(a) Increase of cubic crystal
content from 1 to 100%
1:
1
0
–8 –6 –4 –2 0
intercept
2
Figure 3.24 (a) The closure problem in CSDs, illustrated for cubic grains
(Higgins, 2002a). A straight CSD with a slope of 0.57 (characteristic length
1.73 mm) and an intercept of 8.6 has a volumetric phase proportion of 1%. If
the intercept is increased to 4.0, for example by crystal growth and exponential
increase in nucleation rate, then the crystal content will reach 100% and textural
changes will stop. (b) A fan of straight-line CSDs with 100% crystals is
tangential to a concave-up curve. (c) Straight CSDs can be represented by a
point on a graph of characteristic length (1/slope) against intercept. Slope and
intercept values for materials with 100% crystals proscribe a curve. CSDs can
exist below, but not above this line. The position of the line differs for different
shapes, here indicated by the aspect ratios of rectangular parallelepipeds and
spheres. It is assumed here that the crystals completely fill the volume.
4
74
Grain and crystal sizes
Figure 3.24c shows another view of the same effect in terms of characteristic
length against intercept. All straight CSDs can be defined by their intercept
and characteristic length. Closure limits describe a curve for each crystal
shape. Straight CSDs can only exist below this line.
If a sample has a low crystal content, that is it is far from the closure limit,
then the intercept and slope can change independently giving two degrees of
freedom. However, as the crystal content approaches 100% then any process
that changes the slope of the CSD must also change the intercept – in this
situation there is essentially only one degree of freedom. Also, if the crystal
content is fixed (e.g. Vi ¼ 10%, 50%, 100% etc.) then intercept and slope must
be also linked. Therefore, crystals must dissolve (melt) or divide into subgrains to enable changes in the slope and intercept of the CSD.
3.3 Analytical methods
Grain sizes and size distributions are three-dimensional properties and ideally
they are measured directly. However, as was mentioned in Chapter 2, in many
materials this is either not possible or not the best approach. In those cases data
must be acquired from two-dimensional sections and the true sizes calculated
using stereological methods. Whatever method is used, it is important that
researchers clearly describe their method and make publicly available the basic
data of a study in a useable format. In this way data can be recalculated and
comparison between new and existing studies is made possible.
It is also important to decide and state what material is being measured. This
is particularly important for rocks in which some or all of the crystals are
broken. If broken crystals are included in the CSD then the number of larger
crystals will be reduced and that of smaller crystals increased. The effect will be
much more significant for small crystals if their actual abundance is low or
zero. It is not possible to reconstruct the crystal CSD from a mixed population
of crystals and fragments. A possible solution is to examine each crystal
individually and determine if it is a fragment. If only a few crystals are broken
then a CSD of intact crystals may be determined. If most of the crystals are
broken then there is nothing to be done.
3.3.1 Three-dimensional analytical methods
3.3.1.1 Solid materials
The size of individual grains can be determined by serial sectioning (see
Section 2.2.1). In transparent materials, such as volcanic glasses, the size of
crystals can be measured directly with a regular or confocal microscope (see
3.3 Analytical methods
75
Section 2.5.3). X-ray tomography can be used to determine the size of grains if
adjoining grains can be separated (see Section 2.2.3). The size of crystals is
preserved if they can be extracted intact from rocks by sample disintegration
and solution (see Section 2.4). It is then possible to use techniques developed
for unconsolidated materials (see below).
3.3.1.2 Unconsolidated or disaggregated materials
The size distribution of loose materials may be determined directly by a
number of methods, some of which use unique definitions of size (see below;
Syvitski et al., 1991). The most direct method of grain size analysis is sieving.
Sieves are available in a number of standard sizes and materials. Disposable
nylon sieves eliminate cleaning, which can be laborious. Sieves are not very
efficient for grains less than 500 mm. Ideally the longest axis of the grain passes
vertically through the sieve hole. Under these conditions the maximum size
of grain that can pass through a sieve is equal to an intermediate dimension of
the grain.
The principal use of grain size analysis in sedimentology is to examine
sedimentation patterns. These depend at the first approximation on the settling characteristics of particles. Hence, several analytical methods use controlled sedimentation to measure size (Syvitski et al., 1991). Hence, grain size
is defined by settling velocity and converted to equivalent sphere diameters.
One of the more popular instruments uses X-rays to measure the amount
of sediment that is deposited at the base of the settling tube (Coakley &
Syvitski, 1991).
Another method uses changes in electrical resistance to determine the size
and volume of grains (Milligan & Kranck, 1991). The sample is introduced
into an electrolyte, mixed and then drawn through a narrow aperture. A
constant current passes between the two electrodes on either side of the
aperture. Each particle displaces an equivalent volume of the electrolyte and
hence changes the resistance. If the particles are allowed to pass one at a time,
then each particle will produce a pulse that is proportional in size to the volume
of the particle.
Rapid measurements of grains smaller than 1 mm are possible with a laser
particle diffraction spectrophotometer (Agrawal et al., 1991). The sample is
dispersed in water which is introduced into a transparent cell. Light from one
or two lasers is shone through the cell and allowed to illuminate an array of
concentric circular detectors. Small particles diffract the light more than larger
particles. Hence, the variation of intensity versus deflection angle can be
transformed into the grain size distribution. Measurement is instantaneous
since there is no sedimentation effect.
76
Grain and crystal sizes
3.3.2 Two-dimensional analytical methods
The size of grains can be determined from images acquired from rock surfaces,
slabs, sections and projections, using techniques described in Section 2.2. Such
images can be processed manually or automatically to produce classified
mineral images (see Section 2.3). Finally, the binary images must be processed
to extract crystal outline parameters, such as size.
Grains can be intersected by or projected onto a plane to give a grain outline
(Figure 3.25). Thin sections have a finite thickness and hence small crystals
may be projected whereas larger crystals are intersected. These situations must
be clearly distinguished as the conversions to 3-D data are different. Grain
outline definitions will be discussed first, followed by intersection and projection methods.
3.3.2.1 Grain outline width and length definitions
If a sphere is intersected by a plane, or projected onto one, then the outline
shape is a circle. A circle has a unique, clearly defined length or width: its
diameter. Crystal or bubble outlines generally have irregular shapes; hence
outline width and length are not uniquely defined. The area of each outline is
uniquely defined, but is not the best parameter for stereological calculations of
size (Higgins, 1994). However, in a section the total area of grain intersections
divided by the area measured is equal to the volumetric proportion of the
phase (see Section 2.4.2; Delesse, 1847). Hence, the intersection area is significant and should be measured if possible.
(a) Section view
(b) Upper surface
(c) Thin section
1
2
3
(d) Projection
Figure 3.25 (a) Section view of three opaque crystals in a transparent
matrix. A thin section is indicated in pale grey. (b) The upper surface of the
thin section is imaged. Crystal 1 is not visible. (c) The crystals are imaged in a
thin section. All three crystals are imaged. The outline of crystal 3 is larger
than that on the upper surface. (d) The crystals are viewed in projection.
Crystal 1 presents the same outline as in the thin section. Crystals 2 and 3 have
larger outlines than those in other views.
3.3 Analytical methods
(a) Maximum length
(b) Smallest bounding box
77
(c) Ellipse equal with area
and moment of inertia
(d) Circle of equal area
Figure 3.26 Commonly used definitions of outline size. (a) Greatest length.
(b) Length and width of smallest bounding box. (c) Major and minor axes of
an ellipse of equal area and moment of inertia. (d) Circle of equal area.
Several definitions of outline length and width are possible (Figure 3.26).
1. The length can also be defined as the greatest distance between points on the
outline. This is generally greater than the ‘box length’. This is sometimes known
as the Maximum Feret length.
2. A box can be rotated around the outline until the box width is narrowest or the box
length is greatest – these are commonly the same orientation. The outline length
and width are the dimensions of this box. This is used by some stereological
corrections programs (e.g. CSDCorrections).
3. The outline can be approximated by an ellipse of equal area and moment of inertia.
The outline length and width are the major and minor axes of the ellipse. This is
used by the popular free program NIHImage (see Appendix).
4. The outline can be modelled by a circle of equal area. The outline size is the
diameter of this circle and is commonly called the Feret diameter. This is not a
good choice as it compresses the sizes of outlines.
5. Some programs calculate the width and height of a bounding box with vertical and
horizontal sides (e.g. NIHImage). Clearly this ‘size’ depends on the orientation of
the crystal outline and is not useful.
3.3.3 Intersection mean size methods
3.3.3.1 ASTM mean and maximum size methods
The American Society for Testing Materials (ASTM) has developed standard
procedures for the determination of the apparent grain size of metals from
grain intersection data. The standards only assess apparent mean grain size in
sections and do not make any attempt to extend this to the true 3-D grain size.
78
Grain and crystal sizes
The first standard deals with the problem of estimating the largest grain in a
section, called ALA (as large as) grain size (ASTM, 1992). A further, more
detailed, standard deals with mean grain sizes (ASTM, 1996). The standard
suggests two analytical methods: number of grains per unit area and number
of grain intersections per unit length of test line, with the emphasis on the
latter method. If a metal has a significant orientation fabric then the measurement of three orthogonal sections is advocated. Circular test lines also compensate for anisotropy in the plane of the section. A third standard deals with
mean grain size measurements using semi-automatic or automatic image
analysis methods (ASTM, 1997). Measurements of 400 grains give an uncertainty of about 10% in the mean grain size. It should be remembered that
these procedures do not give grain size distributions, just maximum and
mean grain sizes. Apparent grain sizes are expressed as the ASTM grain
size number, G, where NA is the number of grains per square mm, and NL
is the number of grain intercepts (times entering a grain) per unit length of
test line:
G ¼ 3:32 logðNA Þ 2:95
G ¼ 6:64 logðNL Þ 3:29
3.3.3.2 Grain number by shape independent intersection methods
There are a series of methods that use two parallel sections to determine the
total number of grains per unit volume, NV (Royet, 1991). The advantage of
these methods is that the shape and size of the grains is not important. The
basic method is called the ‘Dissector’.
A ‘reference section’ of area A is defined and grain outlines are determined.
The ‘lookup section’ is parallel to the reference section and separated by
distance H. Grains whose outlines are present in the reference section are
looked for in the lookup section. Q is the number of outlines counted in the
reference section that cannot be identified in the lookup section. Then:
NV ¼ Q=ðAHÞ
The resolution of the method is limited by the spacing of the reference and
lookup sections, and by the ability to trace grains from one section to the
other. It should be emphasised that this method gives the total number of
grains per unit volume and not the grain size distribution. The mean volume of
the grains can be determined if the total volume of the grains is known (see
Section 2.4.2). The mean size of the grains can be calculated if a grain shape is
assumed.
79
3.3 Analytical methods
3.3.3.3 Mean grain size by shape independent intersection methods
The ratio of the surface area to the volume, SV, can be accurately determined
without any assumptions of grain shape or orientation (see Section 2.4.2). This
parameter is a measure of the mean curvature of the grains and has the
dimensions of L1. The inverse of this parameter is a measure of the mean
size of the grains, which is just the mean intercept length. This measure of mean
size is used in some materials science literature (Brandon & Kaplan, 1999).
3.3.4 Extraction of grain size distributions from grain intersection data
3.3.4.1 Nature of the problem
The problem of conversion of two-dimensional intersection or projection size
data to three-dimensional CSDs is not simple for objects more geometrically
complex than a sphere and there is no unique solution for real data (Figure 3.27).
Cube 1:1:1
Prism 1:1:10
Tablet 1:10:10
Tablet 1:2:5
Figure 3.27 Intersections of random planes with regular geometric figures.
It should be noted that tablets tend to give elongated intersections, which are
commonly thought to indicate a lath shape, whereas prisms commonly have a
rectangular cross-section (Higgins, 1994).
80
Grain and crystal sizes
Assume shape?
Yes
No
Direct (unbiased)
methods
Indirect (biased)
methods
Assume distribution?
Yes
Parametric
methods
Figure 3.28
No
Distribution-free
methods
Flow diagram for stereological methods (after Higgins, 2000).
The problem was first attacked in geology by sedimentologists: empirical
solutions were developed early and, despite criticism, have been commonly
applied since then (Friedman, 1958, Johnson, 1994). In the field of igneous
and metamorphic petrology the problem has been discussed by Cashman and
Marsh (1988), Peterson(1996), Sahagian and Proussevitch (1998) and Higgins
(2000). The following treatment generally follows the stereological approach
of Higgins (2000).
Stereological solutions to this problem can be classed into direct or
unbiased methods for which no assumptions about the shape and size distribution of the particles are necessary, and indirect or biased methods, in which
some assumptions are needed (Figure 3.28; Royet, 1991). Indirect methods
can be classified further into parametric methods, where a size distribution
law is assumed, and distribution-free methods that are applicable to all
distributions.
In an ideal world the direct methods would be the best solution. However,
although direct methods for the determination of mean size are very powerful
and simple (Howard & Reed, 1998), they are not available for size distributions. Three-dimensional methods can yield size distributions directly, but
their limitations have already been discussed (see Section 2.2). The lack of a
direct size distribution method has prompted some workers in stereology to
suggest that size distributions should not be determined from sections, or that
size distributions do not give any useful information to practical problems
(Exner, 2004). These limitations, however, seem unduly onerous and hence we
are usually left, therefore, with indirect methods.
3.3 Analytical methods
81
It should be mentioned that the techniques to be discussed below are based
on the assumption that the section measured (or visible part of the surface) is
thin compared with the dimensions of the object. For thin sections viewed in
transmitted light, this assumption means that the crystals must be larger than
0.03 mm, the thickness of a standard thin section. If crystals smaller than this
limit are to be measured, then the crystal outlines are a projection and not a
section, and hence different equations must be used (see Figure 3.25 and
Section 3.3.5).
3.3.4.2 Parametric solutions
The earliest geological methods for stereological conversion were determined
from studies of loose sediments, supplemented by determinations of grain size
from thin sections. However, it was discovered early that there was a systematic bias in thin section data as compared to sieved fractions (Chayes, 1950).
The earliest attempts to solve this problem were based on empirical data:
sediments were analysed by sieving and thin-section analysis (Friedman,
1958). True grain size distributions are commonly approximately lognormal
by mass at the precision level of this type of analysis; hence this method is
actually parametric. The core of the method is to use the quartiles from the
analysis of size in thin sections and transform these to quartiles equivalent to
those of the sieved data using a simple linear equation. This method has been
corrected and modified by others, but was most severely criticised by Johnson
(1994). He pointed out that the method does not work very well for the smaller
size fractions, but only suggested a way of calculating the mean particle size
from intersection measurements.
Kong et al. (2005) proposed another parametric solution for spheres with
lognormal or gamma distributions. Despite the complexity of the mathematical arguments, this method is very restrictive in terms of shape and distribution: it will be shown below that there are much simpler and more powerful
solutions to this problem. They also concluded that a simple factor can be used
to convert 2-D mean sizes to 3-D mean sizes. Again, a much better solution is
to use stereologically exact global parameters (see Section 2.5.2) as there is no
need to assume shape or distribution.
Peterson (1996) proposed a parametric solution based on another distribution: he initially assumed that rock CSDs have a strict logarithmic variation in
population density for a linear variation in size. This distribution is indicated
by the theoretical studies of Marsh (1988b) for simple igneous systems.
Peterson first corrected the data for the intersection-probability effect (see
below). He then applied a further correction using three empirical parameters
that depend on the shape and spatial orientation of the crystals. He found a
82
Grain and crystal sizes
good correlation between the CSDs determined by his methods and the true
CSDs for synthetic data with logarithmic-linear CSDs. However, many
published natural CSDs are not linear. Peterson was aware of this problem
and applied the corrections derived from linear approximations to the
actual curved CSDs. However, he did not verify whether the results for
curved CSDs were accurate using synthetic or natural data. Hence, it is
not clear if these parametric methods can be applied to many natural
systems. It also seems unwise to use a technique to find CSDs that assumes a
CSD shape beforehand. Another limitation of these methods is that they can
only be applied to isotropic (massive) fabrics, unless other parameters are
introduced.
3.3.4.3 Distribution-free solutions
The intersection-probability and cut-section effects are relevant for distributionfree solutions (Underwood, 1970, Royet, 1991). The intersection-probability
effect is that for a population of grains with different true sizes (polydisperse),
smaller grains are less likely to be intersected by a plane than larger grains. The
cut-section effect is that the intersection plane never cuts exactly through the
centre of each particle; hence, even in a population of equal-sized particles
(monodisperse), intersection sizes have a broad range about the modal value,
from zero to the greatest 3-D length. Both problems compound to make the
stereological conversion complex.
The intersection-probability effect is easily resolved for a monodisperse
collection of spheres (Royet, 1991):
nV ¼
nA
D
This equation can be modified so that it applies to polydisperse (many true
sizes) collections of spheres if many size ranges are treated separately: for
instance 0–1 mm, 1–2 mm, 2–3 mm. Other shapes can be accommodated if a
linear measure of the size of the particle is used instead of the diameter.
The simplest formulation of the intersection-probability effect for nonspherical objects in a size interval j is:
nV ðLj Þ ¼
nA ðlj Þ
j
H
j is defined as the mean
Mean projected height (or mean calliper diameter) H
height of the shadow of the particle for all possible orientations (Sahagian &
Proussevitch, 1998). It is commonly close to the other definitions of size
discussed in Section 3.1.2.
3.3 Analytical methods
83
Table 3.1 Some symbols that are used below. Lower case letters
commonly refer to intersections and upper case to 3-D parameters.
Other symbols are discussed later.
A size interval
Size limits of an interval
3-D size
Width of interval j
Mean size in interval j
Sphere diameter
Volumetric phase proportion
Intersection length
Mean intersection length in interval j
Intersection width
Mean intersection width in interval j
Probability that a crystal with a true size in size interval A will
have an intersection that falls in intersection size interval B
j
Mean projected height for size interval j
H
Total number per unit area
nA
Total number per unit volume
nV
nA (lj) Number per unit area for size interval j
nV (Lj) Number per unit volume for size interval j
nV (0) Nucleation density (intercept on size axis)
n0V (L) Population density at length L
Shape parameters
S
Short axis of parallelepiped or minor axis of ellipsoid
I
Intermediate axis of parallelepiped or intermediate axis of
ellipsoid
L
Long axis of parallelepiped or major axis of ellipsoid
j
x, y
L
Wj
Lj
D
V
l
lj
w
wj
PAB
The intersection probability effect uses the assumption that each intersection passes exactly through the centre of the object along the longest axis. This
is very unlikely and hence we have to contend also with the cut-section effect.
The cut-section effect can be solved analytically only for spheres (e.g.,
Royet, 1991). The size distributions of intersections between a random plane
and a sphere of unit diameter rises to a maximum near the diameter of the
sphere (Figure 3.29a). Hence, the most likely intersection diameter is close to
the maximum and the true diameter.
The intersection of a triaxial ellipsoid with a plane produces an ellipse. The
distribution of the major and minor axes (¼ length and width) of the intersection can be evaluated numerically (Figure 3.29). The intersection lengths of
oblate ellipsoids (Figure 3.29a) and intersection widths of prolate ellipsoids
(Figure 3.29d) have size distributions that resemble spheres. Intersection
lengths of prolate ellipsoids (Figure 3.29b) also rise to a maximum at the
84
Grain and crystal sizes
(b) Prolate
ellipsoids
frequency
(a) Oblate ellipsoids
+ spheres
Sphere
1:2:2
1:5:5
1:10:10
1:1:2
1:1:10
1:1:5
0
0.5
1.0
1.5 0
0.5
1.0
intersection length / intermediate axis
(c) Oblate ellipsoids
+ spheres
1.5
2.0
2.5
3.0
(d) Prolate
ellipsoids
frequency
Sphere
1:2:2
1:5:5
1:10:10
0
0.5
1.0
1.5
2.0
2.5
3.0 0
0.5
intersection width / minor axis
1:1:10
1:1:2 1:1:5
1.0
1.5
Figure 3.29 Intersections between a random plane and a triaxial ellipsoid
(this study). One million randomly oriented intersections were calculated for
different shapes of ellipsoids. Intersection lengths and widths are normalised
to true (3-D) intermediate and minor axes of the ellipsoid. Oblate ellipsoids
are short and wide and prolate ellipsoids are tall and thin.
intermediate axis value, but continue to larger sizes. Intersection widths of
oblate ellipsoids rise to a maximum for the minor axis and also continue to
larger sizes (Figure 3.29c).
The distribution of intersection dimensions for shapes such as parallelepipeds must be derived numerically and can be very complex with several modes
(Figure 3.30; Higgins, 1994, Sahagian & Proussevitch, 1998).
As noted above, for spheres and near-equant forms, the mean intersection
length is close to the true 3-D size of the object. However, for monodisperse
populations of randomly oriented anisotropic figures, such as parallelepipeds,
the most-likely intersection length is close to the intermediate dimension
(Higgins, 1994). That is, for a particle 1 2 10 mm, the most-likely intersection length is 2 mm. The most-likely intersection width is close to the short
dimension. The same argument can also be applied to populations of crystals
85
3.3 Analytical methods
1:1:1
1:2:2
1:5:5
1:10:10
1:1:2
1:1:5
1:1:10
(b) Prisms
frequency
(a) Cube
and tablets
0
0.5
1.0
1.5
2.0
2.5 0
0.5
1.0
intersection length / intermediate dimension (l/I)
2.0
2.5
(d) Prisms
frequency
(c) Cube and
tablets
1.5
0
0.5
1.0
1.5
2.0
2.5 0
0.5
1.0
intersection width / short dimension (w/S)
1.5
2.0
2.5
Figure 3.30 Intersections between a random plane and a regular
parallelepiped. Intersection lengths and widths normalised to intermediate
and short parallelepiped dimensions (Higgins, 1994, Higgins, 2000).
with different true sizes. These modal values for intersection length and width
can be corrected to the true length of the crystal, or any other size parameter, if
the shape of the solid is known (see Chapter 4). However, there is another
problem: intersection lengths and widths for a monodisperse population tail
out to smaller and larger values respectively around the most-likely intersection
length. If the simple conversion equations are used, then the effects of tailing
become very important. This is because small intersections of large objects
(corners) will be converted using their apparent length and not their true length.
3.3.4.4 Saltikov method
Saltikov (1967) proposed a method3 of unfolding a population of intersection
lengths into the true length using a function of the intersection lengths
(Figure 3.31). For a polydisperse population, the 3-D distribution of lengths
86
Grain and crystal sizes
3-D
2-D
2-D cross-section
from 1st 3-D class
accumulation from
all larger classes
from 2nd
from 3rd
from 4th
2-D cumulative
distribution
3-D real
distribution
# 1
2
3
4 5 6 7 ...
size classes
Figure 3.31 Cut-section effect. Diagram after Sahagian and Proussevitch (1998).
can be found by applying the function for a monodisperse population of the
same shape to the 2-D length distribution. The values of nV ðLXY Þ for a series of
bins from 1, 2, 3, 4 . . . can be calculated sequentially using the equations of
Sahagian and Proussevitch (1998) as modified by Higgins (2000):
nV ðL1 Þ ¼
nV ðL2 Þ ¼
nV ðL3 Þ ¼
nV ðL4 Þ ¼
nA ðl1 Þ
1
P11 H
1
nA ðl2 Þ nV ðL1 ÞP12 H
P22 H2
2 nV ðL1 ÞP13 H
1
nA ðl3 Þ nV ðL2 ÞP23 H
3
P33 H
3 nV ðL2 ÞP24 H
2 nV ðL1 ÞP14 H
1
nA ðl4 Þ nV ðL3 ÞP34 H
4
P44 H
where subscript 1 refers to the first (largest) size interval.
87
3.3 Analytical methods
The size intervals (bins) are logarithmic as this simplifies the calculations.
However, programs such as CSDCorrections can accommodate any size bins.
Both Saltikov (1967) and Sahagian and Proussevitch (1998) proposed ten size
bins per decade, that is each bin is 100.1 smaller than its neighbour. Higgins
(2000) suggests 5 bins per decade as a starting value as this reduces the total
number of bins and hence the accumulation of uncertainty. It also yields larger
numbers of crystals, and smaller counting errors, in each bin where applied to
typical geological samples. The probabilities, PAB, can be calculated from
numerical models of different crystal shapes (Higgins, 1994, Sahagian &
Proussevitch, 1998). For isotropic fabrics, the shape is constructed mathematically and sectioned using randomly oriented planes placed random distances
from the centre of the crystal model. The mean projected heights also can be
calculated using the same models.
The Saltikov method works well for spheres and near equant objects (spheroids) because the modal 2-D length lies close to the maximum 3-D length (e.g.,
Armienti et al., 1994, Sahagian & Proussevitch, 1998). However, Higgins
(2000) has shown that this method is less successful for complex objects
because the largest 2-D intersections are not very common. That is, the
modal 2-D length is much less than the maximum 2-D length. Therefore, the
less abundant, and hence less accurately determined, larger intersections are
used to correct for the most abundant intersections, introducing large uncertainties in the corrected 3-D length distributions (Figure 3.32a). The problem
is actually worse than might be expected from the works of Saltikov (1967) and
(a) Saltikov method
(b) Modified by Higgins
Per cent in each bin
26
3
0
0.5
1.0
intersection length (l)
1.5
4 2 8
8
13
48
0
6
Oversize
33
frequency
6 8
Undersize
63 4 3 4 5
Undersize
frequency
Per cent in each bin
0.5
1.0
intersection length (l)
1.5
Figure 3.32 The Saltikov method for a tablet 1:5:5. (a) The Saltikov (1967)
method uses the largest interval (shaded) for correcting the smaller
intervals. For a 1:5:5 tablet this interval has only 3% of the intersections
and is hence imprecise. (b) Higgins (2000) modified the Saltikov method by
using a much wider bin that contains 48% of the intersections and is hence
more precise.
88
Grain and crystal sizes
Sahagian and Proussevitch (1998) as both authors placed the maximum intersection length at the upper limit of the size bin. Because true particle sizes can
be distributed throughout the size bin, at the very least the maximum intersection size should be placed at the centre of the bin.
Sahagian and Proussevitch (1998) reduced this problem by ignoring the
class of largest intersections and shifting the probabilities to lower classes.
Higgins (2000) used the most likely intersection length or width (modal value)
to correct for the tailing to other intersections (Figure 3.32b). However, it is
not possible to correct tailing in both directions using this technique – tailing to
intersections either larger or smaller than the modal value must be ignored. In
general tailing to smaller intersections is a much more important problem than
tailing to larger intersections for most CSDs. Hence, the first size interval is
centred on the mode of the intersection length or width.
Some workers have considered the Saltikov technique to be impractical
because of the accumulation of terms, and hence uncertainties, in the smaller
bins. This problem is not always very serious as many of the terms are not
significant, especially for equant and spherical forms. It can also be reduced by
using fewer, wider bins. Intersections larger than the first interval cannot be
corrected precisely with this method. However, if they are added to the first
interval, then this source of uncertainty is minimised.
Clearly, the fewer corrections that are needed the greater the accuracy of the
final data. In many cases use of intersection widths instead of lengths reduces
the amount of corrections necessary (Figure 3.33).
Crystal size distribution
steep?
Yes
No
Either cube
or prism?
Cube?
Yes
No
Use length
measurements
No
Yes
Use width
measurements
Figure 3.33 A possible strategy for choosing width or length intersection
measurements (Higgins, 2000).
3.3 Analytical methods
89
Once the mode of the intersection length or width is used instead of the
maximum values, then the intersection length scale no longer corresponds to
the true length scale: that is, Lx 6¼ lx . For a parallelepiped with a short dimension S, an intermediate dimension I and a long dimension L, or an ellipsoid
with the same dimensions, the modal value of the intersection lengths on
randomly oriented planes is I (Higgins, 1994) hence:
Lx ¼ lx L=I
Similarly, the modal value of the intersection widths is S hence:
Lx ¼ wx L=S
Higgins (2000) found that for tablets, where L/I ¼ 1, a much better fit is
obtained to the test data of Peterson (1996) if the mean intersection length is
used rather than L in the above equations. This gives a maximum difference of
20% in the length scale for 1:10:10 tablets and decreases to zero for cubes. The
1 is determined in terms of
Saltikov equations do not need to be changed as H
the true crystal length.
Other refinements have also been made to the Saltikov method. Tuffen
(1998) has pointed out that the inverse of the mean projected height must be
averaged over the true length interval and proposed that:
Zx
1
1
dL
¼
L
H xy x y
y
where x ¼ upper limit of interval and y ¼ lower limit of interval. This can be
integrated to give:
1
lnðx=yÞ
¼
xy
H xy
The original equations of Saltikov (1967) used logarithmic size bins.
Sahagian and Proussevitch (1998) followed this idea, but introduced more
complex equations for linear size bins. However, if the probabilities used in the
Saltikov equations are calculated for each cycle of corrections then the calculations remain simple for any set of bin sizes. This method of calculation has
been implemented in the program CSDCorrections (Higgins, 2000), so that
previously published data can be corrected. Nevertheless, the use of logarithmic size bins is recommended for most studies.
The Saltikov method can only be used in its simplest form for isotropic
fabrics. However, the probabilities of intersection, and the factors for
90
Grain and crystal sizes
converting intersection widths and lengths to true crystal sizes can be readily
modelled for anisotropic fabrics if the fabric parameters are known or can be
estimated. This is done on a dynamic basis in the program CSDCorrections
(Higgins, 2000). The intersection probabilities also depend on the shape of the
crystals. Crystal shapes will be discussed in Chapter 4.
There have been a number of published methods which are modifications of
the Saltikov method, although the authors were not always aware of this at the
time. Pareschi et al. (1990) developed a stereological procedure that they called
‘unfolding’ which closely resembles the Saltikov method. In this method they
reduced the intersection area to a circle of equal area, but otherwise used the
Saltikov technique for spheres. The ‘Stripstar’ technique and program is again
very similar (Heilbronner & Bruhn, 1998). Crystals are assumed to be spheres
and simple corrections are made using a fixed number of fixed-width size
intervals. The weaknesses of fixed-size intervals have been noted above.
Variable-size intervals can be readily accommodated if population density
instead of frequency is used in the final CSD diagrams (see Section 3.3.7). In
this technique, if all intersections from larger spheres are stripped off from
smaller size bins then negative quantities of spheres (antispheres) are permitted,
which seems to be counterproductive. Programs such as CSDCorrections
(Higgins, 2000) can do the same type of simple corrections using spheres, with
a flexible choice in size intervals and without negative quantities of crystals.
3.3.4.5 Uncertainty analysis of the Saltikov method
Inaccuracy in the determination of population densities arises principally from
three sources (Higgins, 2000): the easiest to understand and quantify is the
counting uncertainty. This is taken to be the square root of the number of
intersections within an interval. It is usually only significant for larger size
intervals with fewer than 20 intersections.
The second source of uncertainty is in the value of the probability parameters PAB used in the Saltikov equations. Although these parameters are
defined precisely for fixed convex shapes, crystals in most natural systems have
more irregular and variable shapes. Another source of uncertainty is that
tailing to intersections larger than the modal interval is included in the
modal interval. Hence, it is difficult to estimate the contributions from this
source to the total uncertainty. However, it is easy to calculate the contribution
of the counting uncertainties of other intersection intervals to the total correction of an interval. This source of uncertainty is most important for small size
intervals, where corrections are most significant.
The third source of uncertainty lies in the conversion of intermediate crystal
dimensions (for intersection length measurements) or short crystal dimensions
3.3 Analytical methods
91
(for intersection width measurements) to true crystal lengths. Uncertainties in
the determination of the crystal shape will produce systematic uncertainties in
both the population density and the size distribution.
3.3.4.6 Data conversion methods used in early CSD studies: Wager
and Grey methods
Many early crystal and vesicle size distribution studies used the following
equation to convert intersection size data to CSDs:
nV ðLj Þ ¼ nA ðlj Þ1:5
for each size interval j.
This equation has been ascribed to many different workers, but appears to
have been first used by Wager (1961) to convert total numbers of crystals in
sections to volumetric numbers, without any discussion of its origin or justification for its use. Wager (1961) may have chosen this equation because it is
the simplest way to convert the dimensions of nA ðlj Þ, 1/L2, to that of nV ðLj Þ,
1/L3. This equation is not discussed in general reviews of stereological methods
(e.g. Underwood, 1970, Royet, 1991, Howard & Reed, 1998) or those applied
to geological problems (e.g. Sahagian & Proussevitch, 1998). It is not correct
and the use of this equation should be discontinued.
Published data determined using Wager’s equation are not wasted and can
be recalculated if information on grain shape, grain fabric and orientation of
the section to the fabric are available or can be estimated. The easiest way is
to use tables of intersection data. If these are not available then the
published CSD graphs can be measured and the size intervals and erroneous
population density determined. In some graphs the actual or base 10 logarithm of the erroneous population density is plotted and this must be converted to the natural logarithm of the erroneous population density. The
Wager equation is then applied in reverse to the erroneous population density
of each size interval and the true value of NA recovered. This can then be used
to calculate the true NV, using a stereologically correct method (see sections
above).
Another method of converting intersection dimensions to three-dimensional
data was proposed by Grey (1970). This equation can be applied to each size
interval
nV ðLj Þ ¼
ðpaÞ1=2 nA ðlj Þ
4r
92
Grain and crystal sizes
where a is the aspect ratio (intersection length/intersection width ratio ¼ l/w)
and r is the radius of a circle of area equal to the area of the crystal outline. If
the intersections are assumed to be square this equation can be recast as
pffiffiffiffiffiffiffiffiffiffiffi
pl=w
nV ðLj Þ ¼ pffiffiffiffiffiffiffiffiffiffi nA ðlj Þ
4 lw=p
This can be simplified to
nV ðLj Þ ¼
pa 1
nA ðlj Þ
4 l
j , but
In this form it is clearly similar to the classic equation nV ðLj Þ ¼ nA ðlj Þ=H
crystal sizes will be offset by a factor of pa=4 from the true values.
3.3.5 Extraction of grain size distributions from projected grain data
The stereological corrections for grains measured in projection are generally
much simpler than those for intersections described above. However, it is
essential that the whole object is seen – that is it does not emerge from either
surface. The intersection probability effect does not exist; hence projective
methods are equally sensitive for small and large grains. The cut-section effect
is replaced by the projected outline effect. As before this can be assessed using
two simple models: ellipsoid and parallelepiped. The most common use of
projected images is in sedimentary petrology where rounded grains are common, hence the importance of the ellipsoid model.
For three shapes of rotational ellipsoid no corrections are necessary as the
projected size is unique (Figure 3.34): (1) The projected size of a sphere equals
its true size. (2) The projected length of oblate ellipsoids equals the major axis
of the ellipsoid. (3) The projected width of prolate ellipsoids equals the minor
axis of the ellipsoid. The length of prolate ellipsoids and the width of oblate
ellipsoids are not unique and these measures should be avoided. For triaxial
ellipsoids the outline length is strongly bimodal with peaks at the major and
intermediate axes; the outline width is also bimodal with peaks at the intermediate and short axes. The complexity of these data necessitates more complex data reduction methods. A method similar to that of Saltikov
(Section 3.3.4.4) can also be applied to these models.
The length of the outlines of both tablets and prisms are easy to interpret
(Figure 3.35a): The outline length is close to the actual long dimension of the
parallelepiped. The width of tablets, when normalised to the intermediate
dimension, is widely dispersed and this measurement should be avoided
93
3.3 Analytical methods
(b) Projected prolate
ellipsoids
Projected lengths of prolate
ellipsoids are invariable
(a) Projected oblate
ellipsoids
1:1:2
1:1:5
1:1:10
0
0.2
0.4
0.6
0.8
projected length/major axis
1.0
1:10:10
1:2:2
0
1:5:5
2
4
6
8
projected width/minor axis
10
Projected widths of oblate
ellipsoids are invariable
Figure 3.34 Projections of ellipsoids with random orientations (this study).
Oblate ellipsoids are wide and short; prolate ellipsoids are tall and thin.
Lengths of projections are divided by the major axis of the ellipsoid; widths
are divided by the minor axis of the ellipsoid. Invariable lengths are always
equal to the major axis of the ellipsoid and invariable widths are always equal
to the minor axis of the ellipsoid. Invariable parameters are always the best
choice as no stereological corrections are needed.
(Figure 3.35c). The normalised width of the prisms is simple and identical for
the dimensions: it varies from 1 to 1.5 (Figure 3.35d). The normalised length of
the prisms is widely dispersed and should be avoided (Figure 3.35b). Blocks
with shapes between tablets and prisms have complex distributions and a
method similar to that of Saltikov can be used.
3.3.6 Verification and correction of grain size distributions using
the volumetric phase proportion
Higgins (2002a) has shown that it is easy to verify that CSDs have been
calculated correctly by comparing the volume of a phase present in the rock
calculated using stereologically exact methods (such as phase area proportion;
see Section 2.7) with that calculated from the CSD.
The volumetric proportion of phase, V, is calculated by integration of the
volume of all the crystals:
V¼s
Z1
0
nV ðLÞL3 dL
94
Grain and crystal sizes
(a) Projected cube
and tablets
(b) Projected prisms
1:1:1
1:1:5
1:1:2
1:2:2
1:5:5
1:1:10
1:10:10
0
0.5
1.0
1.5
2.0 0
0.5
projected length / long dimension
(c) Projected cube
and tablets
1:1:1
1.0
1.5
(d) Projected prisms
1:1:2
1:1:5
1:2:2
1:1:10
1:10:10
0
1:5:5
0.5
0.5
1.0
1.5 0
projected width / intermediate dimension
1.0
1.5
Figure 3.35 Projections of parallelepipeds with random orientations (this
study). Projected width and length are the dimensions of the smallest
rectangle that can be fitted around the projected outline.
where s ¼ shape factor ¼ the ratio of the crystal volume to that of a cube of
side L (see below). Size is defined here as the length of the longest axis of the
smallest rectangular parallelepiped that encloses the crystal.
The shape factor, s, is expanded into a more applicable form:
s ¼ ½1 Oð1 p=6ÞIS=L2
where O is the roundness factor, which varies from 0 for rectangular parallelepipeds to 1 for a triaxial ellipsoid (Higgins, 2000). Obviously most crystals
have a more complex shape, but we are concerned here just with a statistical
measure.
For CSDs with discrete size intervals these equations can be modified to
X
V¼s
nV ðLj ÞL3j Wj
3.3 Analytical methods
95
The summation is for all intervals j. Hence the volumes of a phase can always
be calculated from a CSD. It should be remembered that the natural logarithm
of the population density is plotted on a classical CSD diagram.
If CSDs are straight on a classical CSD diagram of ln(population density)
versus size (see below; Marsh, 1988b) the equation can be integrated to
V ¼ 6snV ð0ÞC4
where nV ð0Þ ¼ intercept and C ¼ characteristic length ¼ 1/slope. It should
be emphasised that this equation is for straight CSDs and that volumetric
phase proportions calculated in this way are sensitive to an incorrect choice of
the shape factor.
These equations are a useful way to verify that CSDs have been correctly
converted from two-dimensional data, by comparing the measured value of
the phase proportion with that defined by the CSD. However, caution must be
exercised as the uncertainty in phase proportion estimated from CSDs can be
significant. This is because much of the phase volume is contributed by the
largest crystals. These are not numerous and hence their population density is
not well known.
It is also possible to correct the CSD so that the volume calculated from the
CSD, VCSD, matches the measured phase volume, VM. This process closely
resembles normalising major element analyses to 100%, and similarly must be
used with caution. We assume that error in VCSD arises from a systematic error
in the relationship between the crystal size used in the CSD and the crystal
dimensions. For instance the crystals may be partly convex, hence the simple
ellipsoid or parallelepiped models cannot be applied. These errors can be
corrected if the true size of each crystal is multiplied by a correction factor F:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
F ¼ 3 VM =VCSD
This correction is available in the program CSDCorrections (see Appendix).
3.3.7 Graphical display of grain size distributions
Grain size distributions can be expressed graphically in many different ways.
Such diagrams are useful to show how well a grain size distribution corresponds
to various theoretical size distributions. Some authors have confused the numerical grain size distribution with the graph used to display such distributions
(e.g. Pan, 2001): obviously, the grain size distribution of a population does not
change when different diagrams are used to display it. Various theoretical size
distributions can be verified by graphical or numerical methods (see Section 2.5).
96
Grain and crystal sizes
3.3.7.1 Size limits of measurements
All analytical methods are limited in the size of the grains that can be accurately
measured (see Section 2). The upper size limit of all methods is the number of
large grains that are in the largest size bin. The precision needed will vary with
the goal, but for diagrams that use a logarithm of the size a minimum of 4
crystals is generally sufficient, but not ideal. In general, the measured maximum
grain size can be increased by augmenting the volume or area measured.
The smallest grain that can be measured is equivalent to the limit of quantification in chemical analyses. The grain size distribution below this value is
undefined. It may be necessary to combine different analytical methods to get
an adequate range in measurable sizes (see Section 2). It is very important that
size limits of the analytical technique are discussed in a study. If no crystals are
recorded below a certain size, then it must be made clear if small crystals are
absent or if the cut-off is an artefact of measurement.
3.3.7.2 Semi-logarithmic: ‘CSD diagram’
In igneous petrology the natural logarithm of the population density of
crystals is commonly plotted against linear crystal size, following the initial
work of Marsh (1988b). This diagram has become known as the CSD diagram
(Figure 3.36).
Population density n0V ðLÞ is defined as the number of crystals per unit
volume within a size interval DL as DL tends to zero (Marsh, 1988b). This is
not very practical as the number of crystals is limited. Marsh (1988b) suggested
that it is better to use the cumulative number of crystals greater than size L,
NV ðLÞ. This is a stable, continuously increasing function from which the
population density can be simply derived:
n0V ðLÞ ¼
dNV ðLÞ
dL
This definition can be used if the true size of all crystals is known, which is
the case for large 3-D datasets derived by tomography or serial sectioning.
However, in many CSD studies data are gathered from two-dimensional
sections and the data transformed to discrete intervals of three-dimensional
sizes. We then need a definition of population density based on size intervals.
In a size interval j, the population density, n0V ðLj Þ, is the number density of
crystals in the size interval, nV ðLj Þ, divided by width of the interval, Wj .
n0V ðLj Þ ¼
nV ðLj Þ
Wj
97
3.3 Analytical methods
(a) CSD as histogram
(b) CSD as line graph
Start of CSD (smaller
crystals not measured)
–8
ln (population density) (mm–4)
ln (population density) (mm–4)
–8
–9
–10
–11
–12
–13
–9
–10
Point in CSD
–11
Termination of CSD
(no larger crystals)
–12
–13
–14
–14
0
10
20
30
crystal size (mm)
40
0
10
20
30
crystal size (mm)
40
Figure 3.36 A simple semi-logarithmic ‘CSD diagram’ following Marsh
(1988b). (a) The diagram is actually a histogram as data are integrated over
the width of each size bin. Each size bin has an uncertainty associated with it
indicated by the uncertainty bar. (b) CSDs are commonly drawn as line
graphs, by connecting the centre of the top of each histogram bar. Higgins
(2003) has suggested that the nature of the ends of the CSDs be clearly
indicated: a vertical bar shows that there are no larger (or smaller) crystals.
A circle just indicates a data point. A circle at the end of the CSD indicates
that the termination is just an artefact of measurement as we have not
examined smaller or larger crystals.
Note that the units of population density, n0V ðLÞ, are length4, as the volumetric number density of crystals, nV ðLÞ, has the units of length3 and it is
divided by a length, the interval width Wj .
A CSD is said to be continuous if it has no empty bins. For a continuous
CSD the value of population density will change smoothly for different bin
widths. If gaps (empty bins) appear owing to scarcity of data then either more
data should be acquired or the bin widths should be widened until every bin
has an adequate number of crystals in it. However, there is no intrinsic reason
why all CSDs should be continuous and so we must be able to accommodate
such gaps. Although commonly drawn as line graphs or as a series of unconnected points, CSD diagrams are actually histograms (Figure 3.37). Therefore,
the line type CSDs should not be connected across empty bins. A suggestion
for clarifying this problem is shown in Figure 3.37b.
If the CSD is linear on this diagram then it can be regressed and the intercept
and slope determined. Some regressions use the uncertainty in each point
as weighting and calculate the goodness-of-fit parameter Q (see Section 2.5.2;
e.g. CSDCorrections 1.3).
98
Grain and crystal sizes
(b) CSD as line graph
2
0
0
ln (population density) (mm–4)
ln (population density) (mm–4)
(a) CSD as histogram
2
The width of isolated
bins are clear
–2
–4
–6
No grains
in this
interval
–8
–10
These points should be
connected to indicate that
there are no empty bins
–2
These points cannot
be connected
–4
–6
Suggested way
of indicating
an empty bin(s)
–8
–10
0
2
4
6
8
10
size (mm)
12
14
0
2
4
6
8
10
size (mm)
12
14
Figure 3.37 The problem of ‘empty’ bins in CSD diagrams. (a) If the data
are presented as a histogram, then the width of the empty bin is clear. (b) On
a conventional CSD diagram the width of the empty bin is not clear. It
is suggested here that the widths of terminal (end) and isolated bins are
indicated by a horizontal dashed line ending a vertical bar.
The slope (m) is sometimes transformed into the characteristic size, C, as
follows:
C ¼ 1=m
This parameter is easier to comprehend than the slope, as it has the units of
length. Indeed, for a straight CSD that extends from zero to infinite size, the
characteristic size is equal to the arithmetic mean size of all the crystals.
3.3.7.3 Bi-logarithmic: ‘Fractal’
The mathematical concept of fractional dimensions – fractals – is recent, but
has had a remarkable impact on the natural sciences, particularly in geology
(Mandelbrot, 1982, Turcotte, 1992). Many natural features are thought to be
self-similar: that is, the texture remains the same at all magnifications. This
behaviour can be described by a fractal dimension.
Fractal (self-similar) distributions can be identified using a bi-logarithmic
diagram – for instance log (number of grains larger than r) against log (r),
where r is the size of the grains (Figure 3.38). A similar diagram of ln (population density) against ln (size) can also be used. It is very important to use a wide
range of crystal sizes, otherwise such a fractal behaviour cannot be identified
(Pickering et al., 1995). The slope of the line can be used to establish the fractal
dimension of the distribution. If the line has several well-defined slopes then
the distribution may be described as multifractal.
99
3.3 Analytical methods
log (number of fragments > r )
Granite fragments
Fractal dimension = 2.50
0.01
0.1
1
10
100
log (equivalent length, r ) (mm)
1000
Figure 3.38 An example of a bi-logarithmic diagram used to examine
possible fractal size behaviour. The size parameter (equivalent length, r) is
the cube root of the particle volume. It is equivalent to the side of a cube
of equivalent volume. The points on this diagram could be linked up as
the distribution is continuous, but this has been omitted for clarity. After
Turcotte (1992).
3.3.7.4 Lognormal and loghyperbolic
Lognormal distributions by mass or volume of grain sizes are commonly
proposed in sedimentary petrology (Lewis & McConchie, 1994b). Cashman
and Ferry (1988) proposed that metamorphic rocks also have a lognormal by
number CSD. Such a lognormal by number model has also been proposed for
CSDs of crystals grown at low temperatures in aqueous systems (Eberl et al.,
1998, Kile et al., 2000). It should be mentioned that chemical abundances are
commonly thought to follow a lognormal by number distribution, although
recent analysis suggests that this is rarely true (Reimann & Filzmoser, 2000).
The parameter that has a normal or lognormal size distribution should
always be specified. In sedimentology the parameter is generally mass, but
volume may be used also. In other applications the parameter is the number of
grains. A distribution that is lognormal by volume will not be lognormal by
grain numbers (Jerram, 2001).
Some authors have used linear frequency diagrams to verify lognormal
distribution (Eberl et al., 1998); this is not the best solution, as deviations
100
(b) ‘Classic’ CSD Diagram
cumulative frequency
frequency
(a) Simple
frequency
diagram
ln (population density)
Grain and crystal sizes
10
0
–10
–20
–30
99
95
(c) CDF
diagram
75
50
25
5
1
–40
0.1
1
10
log (size)
100
0
20
40
size
60
80
0.1
1
10
log (size)
100
Figure 3.39 (a) Lognormal grain size distributions are not clearly displayed
on a simple frequency diagram. (b) They are difficult to recognise on a semilogarithmic ‘CSD diagram’. (c) The clearest graph is of the cumulative
distribution function (CDF) against log (size).
from the ideal distribution are not clearly shown, especially for sizes far from
the mean (Figure 3.39). The best way to verify if data have a lognormal
size distribution is to use a graph of cumulative distribution function (CDF)
of volume, mass or number against log (size). Cumulative data can be
transformed using the inverse function of the standard normal cumulative
distribution available in spreadsheet programs (e.g. Excel function
NORMSINV; see Appendix). Sedimentologists generally specify size in phi (F)
units where
F ¼ log2 ðLÞ
and L is in millimetres.
On this diagram a lognormal distribution plots as a straight line. Mixtures of
lognormal distributions can sometimes be identified as two or more straight
line segments. If intersection data are used then it is very important that
conversion to size distribution is done using stereologically correct methods
(see Section 3.3.2; e.g. Higgins, 2000) and that the data extend for several
standard deviations on either side of the mean. Phi diagrams cannot be
recalculated to population density diagrams unless the volume measured or
the total phase volume is documented.
A number of sedimentologists have proposed that grain sizes follow a
loghyperbolic distribution (Christiansen & Hartman, 1991). Normal distributions have the form of a parabola, when plotted as the log (probability density)
against size; loghyperbolic distributions are hyperbolas in the same space.
Whereas a normal or lognormal distribution is defined by two parameters,
the mean and standard deviation, a loghyperbolic distribution needs four
parameters. Very well defined size data are needed to verify if sizes follow a
loghyperbolic distribution. Much sedimentological size data may not warrant
101
3.3 Analytical methods
this level of treatment as the methods used to obtain the data commonly have
systematic biases. There is no particular graphical solution that is amenable to
verify this distribution. Another distribution that has proved popular is the
Weibull distribution (Murthy et al., 2004). This function can have many very
different forms and is popular with modelling of failure in mechanical systems.
Again there are no particular graphs that can linearise this function.
3.3.7.5 Relative frequency and other methods
Many studies express size data as a simple histogram of relative or absolute
number frequency or per cent volume or mass against size (Figure 3.40a). This
class of diagrams has the benefit of simplicity, but few other advantages. The
height of each bar in the histogram clearly depends on the width of the size
interval: if the width is doubled then the number of grains in that interval will
increase also. If size intervals are not constant this can create a false view of the
data. It is also difficult to see variation in frequency for very small or large
crystals. If this type of diagram must be used then population density (frequency / interval width) should replace frequency, so that the height of the
histogram bars will not be affected by the width of the intervals (Figure 3.40).
Data presented on these types of diagram cannot be converted to a classical
CSD diagram without knowledge of the volume of the sample.
(b) Variable size bins
population density
population density
or frequency
(a) Equal size bins
size
size
Figure 3.40 Plot of the same grain population on two simple, linear size
distribution graphs. (a) Equal sized bins are used to express CSDs in many
studies. For equal sized bins, frequency and population density graphs are
identical. Here, the graph conceals much detail for small grains, and there are
empty bins for large grains. (b) If the population density is used then the bin
widths do not have to be constant and they can vary with the quantity of data.
Here logarithmic-sized bins reveal aspects of the distribution concealed by
fixed width bins.
102
Grain and crystal sizes
3.3.8 Overall grain size parameters: moments of number density
Moments are parameters extracted from the size distribution that are used for
describing overall aspects of the grain population. The material is assumed to
be homogeneous and without preferred orientation (fabric). The moments of a
number density distribution, M0, M1 etc., are defined by
Mi ¼
ZL
Li nV ðLÞdL
0
M0 ¼ total number of grains per unit volume
M1 ¼ total length of grains per unit volume
Higher moments must accommodate a shape factor for departures of the
grains from cubes.
M2 ¼ CA* total projected area of grains per unit volume. The shape factor
CA is equal to the mean projected area of the grain divided by L2. This moment
has been misconstrued by some authors as being related to the total surface
area of the grains. The total surface area can be obtained precisely by other
methods without the need for a shape factor (see Section 2.4.2)
M3 ¼ CV* total volume of grains per unit volume. The shape factor CV is
equal to the volume of the grain divided by L3. This moment has been used to
verify that CSDs have been calculated correctly (see Section 3.2.6). Again this
parameter can be obtained precisely by other methods without the need for a
shape factor (see Section 2.4.2).
If the size distribution function is known (variation of number with size)
then the moments can be evaluated in terms of other textural parameters (e.g.
Marsh, 1988b). Moments are not generally very useful, as the same parameters
can commonly be evaluated more precisely with much less work (see
Section 2.4.2).
3.4 Typical applications
The applications chosen here are representative of the diversity of materials
and problems that have been studied so far and extend two previous reviews of
the subject (Cashman, 1990, Marsh, 1998). However, number of studies on any
rock type is very limited and hence much work is still exploratory. The
scientific quality of publications in this field is extremely variable: some
authors continue to use discredited conversion methods even though they
know that they give incorrect results or have not understood the nature of
3.4 Typical applications
103
the conversion problem. Other authors do not specify how they converted
intersection data, and hence their results are of little use.
Many older CSDs were calculated using conversion equations, such as the
‘Wager’ or ‘Kirkpatrick’ method that are now known to be inaccurate. These
data can be recalculated if the shape of the crystals, the nature of the rock
fabric and the orientation of the section to the fabric can be estimated (see
Section 3.3.4.5). If the raw data are available then this is simple, but more
commonly the published graphs must be measured and back calculated (a few
early CSDs have been recalculated in this way in Higgins, 2000). The greatest
error in the ‘Wager’ method is generally in the intercept of the CSD and the
least in the slope. In addition, the abundance of smaller crystals is much less
accurate than that of large crystals. Where necessary, and possible, data
presented here have been recalculated for the more important papers using
stereologically correct methods (Higgins, 2000) and converted to millimetre
units and natural logarithms.
Another possible pitfall in reviewing published data is the units used in
CSDs, which are not always explicitly indicated. In some CSDs the population
density is calculated in terms of cm4, even if the size axis is in millimetres. In
this case:
lnðpopulation density in cm4 Þ ¼ lnðpopulation density in mm4 Þ þ 9:2
In addition some authors use a logarithmic base 10 scale for the population
density, instead of natural logarithms (base e, 2.718 . . .).
CSD data can be presented on many different types of diagram. If none is
specified then it must be assumed that a ‘classical’ CSD diagram of ln (population density) against size was used.
In the available literature growth rates are expressed in a number of units
that are not easy to comprehend. Here are a few useful conversions for a
typical growth rate: 1010 cm s1 ¼ 109 mm s1 ¼ 1 pm s1 ¼ 0.031 mm a1.
The conventional atomic radius of oxygen is about 120 pm, hence this corresponds to one layer of oxygen atoms every four minutes.
3.4.1 Mafic volcanic rocks
3.4.1.1 Plagioclase in a basaltic lava lake
The study by Cashman and Marsh of the Makaopuhi lava lake (Cashman &
Marsh, 1988) was the first geological application of the CSD theory developed
by Randolph and Larson (1971) and modified by Marsh (1988b), and is
probably the most quoted CSD paper. The Makaopuhi lava lake was on the
104
Grain and crystal sizes
flanks of Kilauea Iki, Hawaii, and was filled in 1965 with basalt initially
containing about 4% plagioclase, as well as olivine and augite. The lava lake
was drilled repeatedly during its solidification; hence a time-sequence of
textural development in the upper part was sampled. CSDs of plagioclase,
ilmenite and magnetite were determined from six samples in a single drill hole
that had total crystal contents from 20% to 80%. Cashman and Marsh (1988)
measured the lengths of a large number of crystals, 3000–5000 per section, and
reduced the data using the ‘Wager’ method (see Section 3.3.4.6). Plagioclase
data have been recalculated in Figure 3.41.
The corrected CSDs are slightly more curved than the CSDs published by
Cashman and Marsh (1988), but can still be considered almost straight
(Figure 3.41). Overall, the new CSDs have shapes and relative differences
similar to those of the original data, but the actual population densities are
different. The sample with the lowest crystal content has the steepest slope and
this decreases regularly with increasing crystal content. The intercept of the
regression is constant. Cashman and Marsh (1988) applied the theory of
12
ln (population density) (mm–4)
Plagioclase in basalt:
Makaopuhi lava lake
10
8
6
20%
83%
4
47%
66%
39%
2
0
0.1
0.2
0.3
size (mm)
0.4
0.5
Figure 3.41 Plagioclase in basaltic magma from the Makaopuhi lava lake,
Hawaii (Cashman & Marsh, 1988). Data were recalculated from graphs of the
reduced data in Cashman and Marsh (1988) and Cashman (1986) using a
crystal shape of 1:3:4, block shape and a massive texture. Some data intervals
have been amalgamated to improve precision and fill empty bins. The
intersection data have been cut off at 0.03 mm, as the crystals are dominantly
in projection below this size and CSD is undetermined (see Section 3.3.2,
Figure 3.25). With the new corrections the slope of the CSD was reduced
by a factor of about 1.5 and the intercept increased by a factor of 15.
Crystallisation is shown in per cent.
3.4 Typical applications
105
Marsh (1988b) to these results, although crystallisation in a lava lake is clearly
not a steady-state process. Using the methods of Cashman and Marsh (1988),
and the slope of the corrected CSD, the plagioclase growth rates were
3.5 1010 to 6.5 1010 mm s1 and the nucleation rate was 3.6 105 to
7.2 105 mm3 s1.
It is profitable to look at the changes in the CSD with solidification.
Cashman and Marsh (1988) interpreted their data to indicate that both the
nucleation and growth rates deceased with solidification. The data for growth
rates are much more secure than those for nucleation rates, which depend on
just one early sample. They considered that this was in response to linearly
increasing undercooling, and that each sample developed under constant
conditions. However, increasing undercooling tends to increase both growth
rates and nucleation rates (e.g. Markov, 1995). It is perhaps easier to think of
the rock textures seen here as developing sequentially and hence the samples
represent a sequence of textural development during solidification (Higgins,
1998). Coarsening is an important process in petrology (see Section 3.2.4) and
I have proposed that it may account for some aspects of the textures of these
rocks (Higgins, 1998). However, closed system coarsening cannot account for
the progressive solidification of the system and hence overall growth must also
be important.
3.4.1.2 Plagioclase in basalt flows
Mt Etna is a large volcano that has been erupting porphyritic hawaiites for the
last 200 years. Armienti et al. (1994) examined the CSDs of plagioclase,
olivine, clinopyroxene and oxide minerals in a range of samples from the
1991–3 eruption. They measured crystals using optical and electronic methods
and converted the data using a modified Saltikov technique. All minerals have
strongly curved CSDs, with steep slopes for microlites (down to 30 mm) and
shallow slopes for large crystals. However, linear size intervals were used in
this study, which resulted in many empty bins for large crystals. If the bins are
widened to eliminate this effect then the slope of the right side is much greater
than presented. Armienti et al. (1994) proposed three intervals of crystallisation: the largest crystals nucleated and grew at depth in the magma chamber.
As the magma rose in the conduit further nucleation and growth occurred. The
microlites formed during quenching of the magma at the surface. However,
mixing of two straight CSDs produces a curved CSD similar to that seen here
(see Section 3.2.3.3), hence only two different periods of crystallisation are
necessary. The similarity of the CSDs of this eruption to those of earlier
eruptions was interpreted to mean that the main features of the volcanic feeder
system had not changed recently.
106
Grain and crystal sizes
The a’a and pahoehoe parts of lava flows have both distinctive surfaces and
internal rock textures. Textural work so far on this problem has been somewhat ambiguous. Sato (1995) examined the CSD of two samples, one of each
texture. Unfortunately the samples were from different eruptions.
Nevertheless, he found important differences in their CSDs. The a’a sample
has a steep CSD that did not extend to large sizes, whereas in the pahoehoe
sample the CSD was much more shallow. He attributed this difference to the
initial temperature of the magmas, before degassing and eruption, with the
pahoehoe having a higher initial temperature. However, this does not seem to
fit with the commonly observed transition from pahoehoe to a’a along
some lava flows. Polacci et al. (1999) only looked at crystal number densities
instead of CSDs. They concluded that crystallisation was occurring within
the lava flow itself. Cashman et al. (1999) only examined the total crystal
number by area and concluded that the magmas were well stirred and hence
there were no delays to nucleation. Clearly there is room for more comprehensive studies.
Sections of many thicker basalt flows reveal two parts that are clearly
defined by the joint patterns: at the base is the colonnade, with regular,
commonly vertical, columnar joints; above it lies the entablature with more
closely spaced, complex joints. Jerram et al. (2003) examined the CSDs of the
two components in the Holyoke basalts (Figure 3.42). Both CSDs are straight,
ln (population density)
8
4
0
Colonnade
Entablature
–4
–8
0
1
2
3
size (mm)
Figure 3.42 Holyoke basalt flow (from Jerram et al., 2003). The entablature
is the upper irregular part of the flow and the colonnade is the part with
parallel columnar joints. The kink in the CSDs probably represents a change
in the crystallisation conditions.
107
3.4 Typical applications
except for large sizes (phenocrysts), suggesting a two-stage crystallisation
process. The colonnade has a shallower slope, which can be interpreted as a
longer residence time, if a steady-state model is assumed. Jerram et al. (2003)
calculated residence times of 8 and 11 years for the entablature and colonnade
respectively for a growth rate of 5 1010 mm s1. This gives reasonable
solidification times for the whole flow. The difference between the CSDs of
the entablature and colonnade may also be interpreted in terms of coarsening
(Higgins, 1998): The entablature then represents an earlier texture than the
colonnade, which had more time to coarsen.
3.4.1.3 Plagioclase growth rates in basaltic magmas.
Cashman (1993) reviewed plagioclase growth rates in basaltic systems. Her
study was based on well-known volcanic systems in which the residence time
can be estimated from the history of the volcano or other parameters and
dykes in which cooling parameters can be easily calculated. For samples with
CSDs she used the slope of straight CSDs to get values for the growth rate. In
other samples she relied just on the mean crystal size and total number density.
She also examined experimental results and concluded that nucleation in
natural systems was generally heterogeneous.
Her main conclusion was that growth rates varied with cooling rate, but that
the parameter was not very sensitive (Figure 3.43). A cooling time of 3 years
log (cooling rate) (°C h–1)
2
0
–2
–4
–6
–4
Iritono
log (growth rate) (mm s–1)
Makaopuhi
Mauna Loa
–6
Salvador
Palisades
–8
–10
2
4
6
8
log (crystallisation time) (s)
10
12
Figure 3.43 Plagioclase growth rate against cooling rate or cooling time
(Cashman, 1993).
108
Grain and crystal sizes
(¼108 s ¼ cooling rate of 0.008 8C h1) gives a growth rate of 109 mm s1 and
a cooling time of 300 years (¼1010 s ¼ cooling rate of 0.00008 8C h1) gives a
growth rate of 1010 mm s1. Although some of the CSDs used by Cashman
were measured from sections and not calculated with the appropriate equations, the errors in the slope are small compared to the range in the slope
values. Such growth rates should be only used in a general way to determine
the residence time or other parameters to within an order of magnitude.
It is not clear how these growth rates can be modified for more silica-rich
magmas. In general such magmas are more viscous and hence have lower diffusion rates. However, this is critically dependent on the water content of the
magmas at depth, which is generally poorly known. Cashman (1992) estimated
that the crystallisation rate for plagioclase in dacite was 5–10 times less than in
basalts. Higgins and Roberge (2003) have proposed that crystallisation is not
continuous but cyclic in an andesite magma chamber. If this is generally true then
it may not be possible to determine crystal growth rates from natural systems.
3.4.1.4 Plagioclase in andesite
Mt Taranaki is an active andesite stratovolcano in New Zealand (Higgins,
1996a). Lavas on the flanks of the volcano contain abundant, generally
euhedral plagioclase crystals, comprising up to 30% of the rock on vesiclefree basis. Many crystals have concentric zones of dusty inclusions, which are
commonly thought to represent periods of partial melting or re-heating. The
CSDs define a broad curved band, but each individual CSD is much straighter
than the overall trend suggests. Indeed, individual CSDs are almost linear
down to small crystal sizes, suggesting that nucleation and growth dominated
(Figure 3.44a). If the crystals grow in a steady-state magma chamber with a
growth rate of 1010 mm s1 then characteristic lengths indicate that the earliest lavas (Figure 3.44b; Staircase group) had a residence time of 50 years.
Subsequent lavas in the Castle and Summit groups had slightly longer residence times of 50–75 years. The youngest magmas, from both Egmont Summit
and the Fantham’s peak, a secondary vent on the side of the volcano, have the
shortest residence times of 30 years. Variations in residence time may reflect
changes in the magma chamber shape or depth, or the temperature of the
surrounding rocks.
The Dome Mountain lavas comprise a comagmatic pile of some twenty
flows with a total thickness of about 300 m (Resmini & Marsh, 1995). Most
flows are andesitic, but there is some variation to more mafic and felsic
compositions. Most of the plagioclase CSDs show no sign of crystal fractionation or accumulation, hence they can be used to give magma storage times. If a
growth rate of 109 mm s1 is assumed then most magmas have residence
109
3.4 Typical applications
0.20
(b)
(a) Plagioclase in andesite,
Mt Taranaki, New Zealand
6
4
2
0
–2
0.16
0.14
Summit &
Castle
0.12
0.10
0.08
–4
–6
Staircase
0.18
characteristic length (mm)
ln (population density) (mm–4)
8
Fantham's
Peak
0.06
0
1
2
size (mm)
3
10
20
30
volumetric phase (%)
Figure 3.44 Plagioclase in andesite lavas from Mt Taranaki (Egmont
Volcano), New Zealand. The original data from Higgins (1996a) were
recalculated and published in Higgins (2000). (a) In a conventional CSD
diagram it is not easy to discern variation trends between samples.
(b) Although the CSDs are slightly curved it is still profitable to calculate
the characteristic length (1/slope) and plot this against the dense-rock
volumetric phase per cent of plagioclase, calculated from the total area of
the crystal intersections and corrected for vesicles. Three groups of lavas
clearly stand out on this diagram.
times of 1.5 to 4 years, except for two of the lower lavas with residence times of
4 to 9 years. The residence times correlate inversely with nucleation density
and broadly correlate inversely with silica content: high-residence-time, lowsilica lavas occur at the base of the pile, whereas low-residence-time, high-silica
lavas occur at the top of the pile. One interpretation of the residence time data
is that the lavas are from an open system of more or less constant residence
time but varying spatially in system locations; e.g. a magma chamber in which
steady-state recharge and eruption may have been reached. Alternatively,
these lavas may represent small samples, closely spaced in time, of a slowly
cooling, large batch or reservoir of magma. The greater residence times associated with the lower flows may indicate the increased amount of time necessary to form the conduit of the magma-volcano system by the oldest magma or
simply that part of the system with the highest liquidus temperature.
The Soufriere Hills volcano, Montserrat started to erupt at the end of 1995
after a hiatus of 300 years and continues to this day. The volcano is dominated
by a dome of andesite that sheds pyroclastic flows as it inflates. Typical dome
fragments are glassy and rich in plagioclase crystals – up to 35% for crystals
larger than 0.03 mm. The pyroclastic material is texturally similar, but
110
Grain and crystal sizes
vesicular. A particular aspect of these rocks is the abundance of large amphibole oikocrysts, up to 12 mm long, many of which are loaded with plagioclase
crystals. Oikocrysts seal in early textures and preserve aspects of them for later
examination. Hence, it is possible to see a sequence of textures in such rocks
that can show how the CSD has varied with time (Higgins, 1998). Plagioclase
CSDs in both dome fragments, pyroclastic rocks and within amphiboles were
examined by Higgins and Roberge (2003).
The texture of the material outside the oikocrysts, the matrix, largely preserves the last state of the magma before eruption whereas the oikocrysts
preserve earlier textures. All matrix CSDs are strongly curved and broadly
similar (Figure 3.45). CSDs of plagioclase from the oikocrysts are coincident
with the CSDs of the matrix, but do not extend to the same sizes. In fact, there
appears to be an almost continuous range of maximum crystal sizes in different
matrix samples and oikocrysts. The instability of amphibole at depths above
5 km indicates that plagioclase crystallisation began below that level.
The oikocryst and matrix plagioclase textures cannot be related by simple
growth of plagioclase. Plagioclase crystals must start small and numerous and
progress towards small numbers of larger crystals. The overall process must
(a) Plagioclase in amphibole oikocryst
(c) Plagioclase CSD
I mm
(b) Plagioclase in matrix
ln (population density) (mm–4)
10
8
6
4
Plagioclase in
matrix (groundmass)
2
0
–2
Plagioclase in
amphibole
–4
–6
0
1
size (mm)
2
Figure 3.45 Plagioclase in andesites of the Soufriere Hills volcano,
Montserrat (Higgins & Roberge, 2003). (a) Plagioclase chadocrysts in
the amphibole give an early view of the texture in the magma chamber.
(b) Plagioclase in the matrix gives a view just before the eruption – same scale
as the oikocryst. (c) Plagioclase CSDs in both environments are summarised by
envelopes that represent the range of a number of samples. The two groups of
CSDs are similar for small sizes, but that of the matrix continues the curve to
larger crystals. A similar variation can be seen in the individual matrix CSDs.
111
3.4 Typical applications
size
ln (population density)
size
A further period of
solution and
coarsening flattens
CSD
ln (population density)
size
ln (population density)
size
Further nucleation
and growth.
Oikocrysts develop
trapping this texture
Solution and
coarsening flattens
CSD
ln (population density)
ln (population density)
Initial texture
developed by
nucleation and
growth
Further nucleation
and growth gives
final CSD seen in
erupted material
size
Liquidus
temperature
Buffering
Buffering
Solidus
per cent
liquid
100%
0%
Nucleation
and growth
Solution and
coarsening
Mafic
magma
injection
Nucleation
and growth
Solution and
coarsening
Nucleation
and growth
time
Mafic
magma
injection
Figure 3.46 Soufriere Hills volcano, Montserrat. The solidification model
of Higgins and Roberge (2003) involves repeated cycles of nucleation and
growth, followed by coarsening.
progressively extend the CSD towards larger crystals. Such an overall process
must involve both crystallisation and solution. Higgins and Roberge (2003)
proposed a model that involves repeated cycles of crystal growth and coarsening (Ostwald ripening) in response to varying degrees of undercooling and
superheat (Figure 3.46). The first cycle started when plagioclase crystallised in
response to increasing undercooling in the magma chamber. This phase ended
when the undercooling decreased. This could have been caused by addition of
heat from mafic magma intruded into the magma chamber or by changes in
pressure (depth) or volatile fugacity. Under these conditions coarsening began.
Small crystals dissolved to feed larger crystals. Numerous repetitions of this
cycle will yield the observed complexly zoned large crystals. At Montserrat the
overall process has increased the total crystal content of the magma, but this
may not necessarily happen elsewhere. Finally, small crystals grow in response
to uplift that just precedes the eruption. Amphiboles may have also grown by
the same process.
112
Grain and crystal sizes
3.4.1.5 Olivine in picrite
The 1959 eruption of Kilauea Iki comprised a series of lava flows and created
a lava lake that has only recently completely solidified. The magma was
picritic, a very unusual composition for Hawaii. Mangan (1990) studied the
CSDs of olivine in samples of the pyroclastic deposits of this eruption. The
material is not very easy to measure because some of the olivine crystals are
xenocrystic, and were derived from disaggregated mantle peridotites. Such
crystals have solid-state deformation features which are lacking in the
magmatic olivine. However, this distinction becomes more difficult for
smaller crystals. The magmatic olivine crystallised in a magma chamber
beneath the volcano. Mangan (1990) measured a small number of crystals
from many different samples, but data have been summed here to improve the
precision (Figure 3.47). The turn-down for small crystals may be an artefact of
measurement. Although the CSD of magmatic crystals is slightly curved, a
0
Plagioclase in picrite: 1959 eruption
of Kilauea Iki, Hawaii
ln (population density) (mm–4)
–2
–4
‘Magmatic’ crystals
–6
Xenocrysts
–8
–10
0
1
2
3
size (mm)
4
5
6
Figure 3.47 Olivine in picrite scoria from the 1959 eruption of Kilauea Iki
(Mangan, 1990). Data have been recalculated from the original length
measurements supplied by Dr Mangan using the shape of crystals in
samples from the lava lake that developed later in the eruption that we have
measured in our laboratory. For both populations of crystals the aspect ratios
were 1:1.25:1.25, with a roundness of 0.5 and crystals were randomly
orientated (massive fabric). ‘Xenocrysts’ show evidence of solid-state
deformation; ‘Magmatic’ crystals lack such deformation.
113
3.4 Typical applications
characteristic length of 0.34 mm can be calculated. Mangan (1990) chose
somewhat arbitrarily a growth rate for olivine of 109 mm s1 as this was
what was used for olivine in the Makaopuhi lava lake (Marsh, 1988a). If we
apply the Marsh (1988b) steady-state model, then this gives a residence time of
11 years. The significance of the CSD of the olivine xenocrysts is not clear and
more data are needed.
3.4.1.6 Anorthoclase in phonolite
Dunbar et al. (1994) have looked at anorthoclase phenocrysts in phonolite
from Mt Erebus, Antarctica. These crystals are up to 90 mm long and are very
abundant. They sampled the crystals in volcanic bombs thrown out by the
volcano. These bombs can be disintegrated easily and the crystals separated
and measured using a ruler. The three samples measured all had straight CSDs
down to 10 mm, the measurement limit (Figure 3.48). They did not attempt to
extend this CSD to smaller sizes with a thin section; hence it is not clear if these
rocks lack very small crystals.
Dunbar et al. (1994) concluded that anorthoclase had low nucleation rates
but that it crystallised continuously through time. Anorthoclase should float
in a phonolite magma, but the authors considered that the uniformity of the
three CSDs did not support this idea. However, it seems to me there is no
reason why flotation cannot produce uniform CSDs. The low overall crystal
Anorthoclase in phonolite:
Mt Erebus, Antarctica
ln (population density) (mm–4)
–8
–10
–12
1984-1
1984-2
1988
–14
0
20
40
size (mm)
60
80
Figure 3.48 CSD of anorthoclase from Mt Erebus (Dunbar et al., 1994).
The crystals were extracted whole by disintegration of three volcanic bombs
erupted in 1984 and 1988. The original data have been converted so that size
equals length in mm using a width/length ratio of 0.33.
114
Grain and crystal sizes
abundance and large sizes suggest that coarsening was important. Initially
there may have been a high nucleation rate and many small crystals, but these
were removed during coarsening. A key test would be to look at the abundance
of small crystals – their absence would suggest coarsening. This could well be
combined with accumulation by flotation.
3.4.2 Mafic plutonic rocks
Quantitative textural studies of plutonic rocks really started with Jackson’s
work on the Stillwater complex in 1961 (Jackson, 1961). He measured crystal
size distributions of olivine, orthopyroxene and chromite, but did not record
the area that he measured; hence his data cannot be completely compared with
modern data.
3.4.2.1 Gabbro
The Oman ophiolite is one of the best exposed and preserved sections of the
oceanic crust, and is thought to have formed at a fast or medium spreading
ridge. Garrido et al. (2001) studied plagioclase in the lower and middle sections, whereas Coogan et al. (2002) examined plagioclase in the uppermost
gabbros, just below the sheeted dykes. In both studies plagioclase CSDs were
concave down, with straight right limbs and a clear lack of small crystals. The
simple shape suggests a simple history: Coogan et al. (2002) proposed that a
high-level magma chamber was filled with primitive, aphyric magma and that
all crystals grew there in place. Garrido et al. (2001) considered that the lack of
small crystals reflected coarsening during the early phases of the evolution of
the textures: late reheating to 500 8C did not cause appreciable grain growth.
Other authors have suggested that grain-boundary migration during coarsening can be halted by the presence of other phases (grain-boundary pinning).
However, here it is shown that this process did not significantly limit the
coarsening process. Garrido et al. (2001) divided up their gabbro section into
two parts on the basis of the CSDs and interpreted this in terms of near-vertical
isotherms close to the ridge axis.
The Kiglapait intrusion is one of the younger components of the Nain
Plutonic Suite. It has a conical form about 30 km in diameter, and is about
8.5 km thick in the centre (Morse, 1969). It has been well studied because it is
essentially undeformed and unaltered. The lower zone is composed of plagioclase and olivine; clinopyroxene joins the assemblage towards the top and
marks the base of the upper zone.
Higgins (2002b) determined the CSDs of plagioclase, olivine clinopyroxene in a series of 13 sections from the base to the top of the intrusion
115
3.4 Typical applications
2
2
0
ln (population density) (mm–4)
ln (population density) (mm–4)
0
–4
–8
–2
–4
10
20
–6
–8
(b) Olivine
0
–2
–4
–6
–8
(a) Plagioclase
–10
–10
0
2
4
size (mm)
6
8
0
2
4
6
size (mm)
ln (population density) (mm–4)
0
(c) Pyroxene
–2
–4
–6
Kiglapait intrusion,
Labrador, Canada
–8
–10
–12
–14
–16
0
5
10
15
size (mm)
20
25
Figure 3.49 CSDs of plagioclase, olivine and clinopyroxene from the
Kiglapait intrusion, Canada (Higgins, 2002b).
(Figure 3.49). All CSDs lack small crystals and are slightly curved to the right,
concave up. However, the curvature is not as strong as first appears: the CSDs
define a fan, as expected if there is little variation in mineral abundance
(Higgins, 2002a). Most of the textural variation is the result of variable degrees
of coarsening, following the Communicating Neighbours model (DeHoff,
1991). The degree of coarsening is lowest at the base, when the magma was
first introduced into cool host rocks. Coarsening reaches a maximum when
the intrusion was 20% solidified, and then decreases towards the top of the
intrusion. This reflects the lower volume and hence heat production of the
crystallising magma, which could be therefore dissipated more readily. There
does not seem to be a correlation between the degree of coarsening of the
olivine and plagioclase, suggesting that they were not always on the liquidus at
the same time.
116
Grain and crystal sizes
Part of the lower zone is well layered, in some cases with well developed
mineral grading. If such layers were produced by crystal settling then we would
expect to see grading in both the amount of the phases and their size distributions. One of these layers was examined in detail. A settling origin is strongly
suggested by the sharp base rich in olivine, which grades up into a top rich in
plagioclase. However, plagioclase and olivine CSDs from the base and top do
not show the expected effect. This suggests that subsequent coarsening has
obscured the grain size effects of crystal setting.
3.4.2.2 Troctolite and anorthosite
Higgins (1998) showed that the textures of plagioclase crystals within large
olivine oikocrysts preserve a textural sequence of the formation of anorthosite.
Such textures have been quantified in a troctolite/anorthosite from the
Proterozoic Lac-Saint-Jean anorthosite suite, Canada (Figure 3.50).
Plagioclase CSDs indicate that it initially nucleated and grew in an environment of linearly increasing undercooling, producing a straight-line CSD.
During this phase, latent heat of crystallisation was largely removed by circulation of magma through the porous crystal mush. By about 25% solidification the crystallinity was such that it reduced, but did not eliminate, the
circulation of magma, resulting in the retention of more latent heat within
the crystal pile. The temperature increased until it was buffered by the solution
of plagioclase close to the liquidus temperature of plagioclase. Nucleation of
plagioclase was inhibited and conditions were suitable for coarsening of both
plagioclase and olivine to occur.
Higgins (1998) considered that the shapes of the plagioclase CSDs fit better
the Communicating Neighbours equation of coarsening, rather than the classic Lifshitz-Slyozov-Wagner equation (see Section 3.2.4), as do other examples
drawn from the literature. If olivine started to nucleate at a higher temperature
than plagioclase, then during the coarsening phase olivine would have been
more undercooled than plagioclase, and would have had a higher maximum
growth rate. Under these conditions olivine would coarsen more rapidly than
plagioclase and engulf it. Hence the order of crystallisation determined from
the textures would be the reverse of the order of first nucleation of the two
phases, from equilibrium phase diagrams. Maintenance of the temperature
near the plagioclase liquidus may also inhibit the nucleation and growth of
other phases.
The upper part of the unmetamorphosed Sept Iles intrusive suite contains a
series of anorthosites about 1 km thick (Higgins, 1991, Higgins, 2005). Parts of
the anorthosite are very well foliated with abrupt transitions to massive
anorthosite. In an early study I examined quantitatively the textures of one
117
3.4 Typical applications
(a) Plagioclase in olivine oikocryst
40%
crystallised
(b) Plagioclase in matrix
50%
crystallised
80%
crystallised
40%
crystallised
50%
crystallised
80%
crystallised
10 mm
10 mm
ln (population density) (mm–4)
0
(c) Plagioclase CSDs
40%
–2
–4
–6
Oikocrystic anorthosite,
Alma, Lac-Saint-Jean
anorthosite suite
50%
80%
–8
Matrix
100%
–10
–12
–14
0
5
10
15
size (mm)
20
25
Figure 3.50 Plagioclase in oikocrystic anorthosite (troctolite) from the LacSaint-Jean anorthositic suite (Higgins, 1998). (a) A large olivine oikocryst
was subdivided into regions of varying plagioclase content: 40, 50 and 80%.
They are considered to preserve a textural development sequence. (b) The
anorthosite from between the oikocrysts is termed the matrix and is 100%
plagioclase. (c) Plagioclase CSDs from the oikocryst regions and the matrix.
Data were recalculated from the original data using a shape of 1:3:3 for the
oikocryst samples and 1:1:1 for the matrix sample.
sample from each facies with the goal of determining the origin of the lamination (Figure 3.51; Higgins, 1991). One aspect that is particularly striking is the
strong difference in crystal shape between the thin tablets of the laminated
anorthosite (1:5:7) and the stubbier crystals of the massive anorthosite
(1:2.5:4). The recalculated CSDs for the laminated and massive anorthosite
are essentially collinear, but the massive anorthosite extends to larger sizes.
There are no crystals smaller than 1 mm for the massive anorthosite and 2 mm
for the laminated anorthosite. In the figure the CSDs terminate at 2–3 mm
118
Grain and crystal sizes
–2
ln (population density) (mm–4)
Plagioclase in anorthosite:
Sept Iles, Canada
–4
–6
–8
–10
Massive
–12
Laminated – Section normal to lamination
Laminated – Section parallel to lamination
–14
0
5
10
size (mm)
15
20
Figure 3.51 Plagioclase in laminated and massive anorthosite from the Sept
Iles intrusive suite (Higgins, 1991). The original data have been recalculated
using the methods of Higgins (2000) and a shape of 1:2.5:4 for the massive
sample and 1:5:7 for the laminated sample. Two orthogonal sections of the
laminated sample were analysed: It is gratifying that they give very similar
CSDs as there is, of course, only one actual CSD for a phase in a rock sample.
because of analytical difficulties. The similarity of the CSDs indicates that the
anorthosite of the two facies crystallised in the same pressure, temperature
and time environment, but that other processes were responsible for producing
the foliation and differences in crystal shape. The most likely candidate is
simple shear which can produce the foliation and also remove magma
around the growing crystal depleted in the plagioclase component (see
Section 5.2.2.2).
3.4.2.3 Impact melt rocks
The Sudbury igneous complex is the melt sheet produced by a large meteorite
impact. The overall composition is a norite, but the body is differentiated. It
may have been superheated when first produced. Zieg and Marsh (2002) have
reported CSDs from this fascinating unit, but unfortunately give few details.
They state that the plagioclase CSDs are linear, but do not present any actual
CSDs. It is also not clear if small crystals are absent, as in many plutonic rocks.
They have shown however, that the characteristic length of the crystals
increases systematically with height in the melt sheet by a factor of two.
They used these data and a simple cooling model to calculate an effective growth
rate of 1.5 1013 mm s1 and a cooling time of 1.8 1012 s (57 ka). These
3.4 Typical applications
119
values are approximately collinear with the data obtained from volcanic rocks
by Cashman (Figure 3.43; 1993).
3.4.3 Silicic volcanic rocks
3.4.3.1 Plagioclase in dacite
Phenocrysts and microlites of plagioclase from the 1980–6 eruptions of
Mt St Helens, USA, were examined by Cashman (1988, 1992). Both phenocrysts
and microlites have straight CSDs and appear to have similar slopes, although
this is difficult to verify as different measurement techniques were used in the
two studies. The CSDs of microlites were combined with geophysical estimates
of the residence time to calculate growth and nucleation rates. Cashman
considered that the growth rate only varied by a small amount for different
values of undercooling, but that the nucleation rate was strongly affected. This
confirms standard nucleation models. She proposed a two-stage crystallisation
model in which most of the phenocrysts nucleated at depth. At the end of the
initial part of the eruption the magma resided in a chamber at a depth of
4–5 km. Further growth in this chamber was caused by coarsening, rather than
nucleation of new crystals.
The Kameni volcano is part of the Thera (Santorini) volcanic complex, an
active volcano in the eastern Mediterranean (Higgins, 1996b, Druitt et al.,
1999). The dacite lavas of the Kameni islands have been erupted over the last
2000 years and form an unusually useful field area: they are fresh, easily
accessible and well-dated. All samples are porphyritic: the most common
megacryst phase is plagioclase (typical mode 12%), with clinopyroxene, orthopyroxene, magnetite, olivine and apatite, all set in a fine-grained to glassy
matrix with plagioclase microlites.
None of the plagioclase CSDs are linear, but curved concave upwards
(Figure 3.52: Higgins, 1996b). Hence, a simple model based on a single
population of crystals, such as that of Marsh (1988b), cannot be applied.
Higgins (1996b) proposed that such CSDs could be produced by mixing of
two populations of crystals, each with overlapping linear CSDs, termed microlites and megacrysts. The two populations can then be examined separately.
Application of the Marsh model (1988b) with growth rates of 1010 mm s1
gives residence times of 6–13 years for the microlites and 24–96 years for the
megacrysts. These residence times broadly accord with those determined from
diffusion rates in plagioclase (Zellmer et al., 2000). If the residence times of the
megacryst populations are projected back from the eruption times, then some
of the times of emplacement of the megacryst magmas are found to coincide
120
Grain and crystal sizes
10
Plagioclase in dacite,
Kameni, Greece
ln (population density) (mm–4)
8
6
4
2
0
–2
–4
0
0.2
0.4
0.6
0.8
size (mm)
1.0
1.2
1.4
Figure 3.52 Plagioclase in dacites from Kameni volcano, Greece (Higgins,
1996b). Data have been reconverted using a massive block crystal shape with
an aspect ratio of 1:4:4. These concave up CSDs can be modelled by a mixture
of two populations of crystals, each with a straight CSD.
with the dates of earlier eruptions. This suggests that the megacrysts are ‘leftovers’ of earlier injections of new magma into a shallow chamber: that is some
magma remains after each eruption and continues to crystallise. Each eruption
is initiated by the emplacement of new magma with few or no crystals, which
mixes with the older magma. Eruption followed 6–13 years after mixing. Such
a model would suggest that some porphyritic magmas are products of a
shallow magma chamber that is never completely emptied, just topped up
from time to time.
Turner et al. (2003) examined basaltic andesites using CSDs and uranium
disequilibrium and found a profound disagreement in the residence times
determined by these two methods. They suggest that this is caused by incorporation of existing cumulates into new magmas, similar to the processes
proposed for the Kameni dacites.
3.4.3.2 Microlites
Microlites occur in many volcanic rocks and their growth may trigger eruptions. For stereological reasons prismatic microlites are not easy to measure in
section although they can be measured easily in projection. Castro et al. (2003)
used optical scanning to measure the CSDs of pyroxene microlite prisms in
obsidian from Inyo Domes, California. On a classical CSD diagram the CSDs
are straight from 5 to 30 mm, with a turn-down for smaller sizes. This is not an
3.4 Typical applications
121
artefact, as the lower size limit is 0.5 mm. On a simple frequency diagram the
distributions appear to be lognormal. Such distributions were interpreted to
reflect a short nucleation event, followed by growth.
Hammer et al. (1999) examined plagioclase microlites in early, pre-climatic
dacite tephras from the 1991 Mt Pinatubo eruption. This material is particularly difficult to measure as the microlites are small and have shapes that vary
strongly with size: the smaller crystals are tabular, whereas the larger crystals
are prismatic. Nevertheless CSDs from 13 different surge-producing blasts
were all straight. The slope was used to characterise the degree of plagioclase
supersaturation: it was found to vary with the repose time between the eruptions
(40 to >170 minutes). Overall crystallisation passed from a nucleationdominated regime, with tabular crystals and steep CSDs to a growthdominated regime with shallow CSDs and prismatic crystals.
3.4.3.3 Zircon in rhyolite tuffs and lavas
The CSDs of minerals in explosively erupted rocks are of great interest, as
they can tell us a lot about the state of the magma chamber just before the
eruption. However, crystals are commonly broken by the force of the eruption;
hence there have been few studies to date. One way around this problem is to
look at zircon and quartz crystals, which are generally robust enough to
survive intact.
Bindeman (2003) liberated zircon and quartz crystals from the Bishop Tuff
and other silicic ash-flow tuffs by crushing and acid digestion. The crystals
were then measured under a microscope. This technique avoids problems
introduced during stereological conversions and enables the measurement of
a large number of crystals with a very wide range in size. Both quartz and
zircon CSDs are concave down on a classical CSD diagram, with a straight
right-hand side (Figure 3.53). However, the distributions are lognormal when
plotted on a CDF diagram, and this is confirmed by other statistical tests.
Bindeman (2003) interpreted these distributions using the single-stage
size-dependent growth model of Eberl et al. (2002). However, as Bindeman
(2003) points out, both zircon and quartz have complex zoning patterns,
indicating periods of growth and solution. It seems more likely that the
lognormal distribution is a result of the interplay of these processes rather
than the simple model of Eberl et al. (2002), which in any case is not supported
by theory. The concave down shape of the CSDs is found in other coarsened
(ripened) rocks; hence coarsening may well dominate here, especially just
before eruption.
CSDs of zircon in rhyolite lavas closely resemble those from tuffs described
above: they are concave down and appear to have a lognormal distribution
122
Grain and crystal sizes
ln (population density) (mm–4)
Bishop Tuff – zircon
0
Early magma
Late magma
–2
–4
0
100
200
size, long axis (μm)
300
Figure 3.53 Zircon CSDs in the rhyolitic Bishop Tuff, California (from
Bindeman, 2003). One sample from the early part of the eruption (top of
the magma chamber) did not differ significantly from three samples from the
later part of the eruption (base of the magma chamber).
(Bindeman & Valley, 2001). Bindeman and Valley (2001) used the slope of
the right-hand side of the CSD to calculate a residence time of 100–200 ka for
these zircons; however, as they point out, the steady-state CSD model may not
be applicable here, the growth rate that they used is not well established and
the zircons may be xenocrysts. Instead they suggest that coarsening may be
important.
3.4.4 Silicic plutonic rocks
3.4.4.1 Orthoclase, plagioclase and quartz in granite porphyry
Mock et al. (2003) have done the first comprehensive quantitative textural
study (size, shape, orientation and position) of a shallow level porphyritic
pluton, using a vertical drill core that intersects the entire pluton. They
examined the CSDs of orthoclase, quartz and plagioclase in slabs and found
that all these minerals have remarkably straight CSDs for crystals larger than
1 mm. The abundance of crystals smaller than this limit is not clear. There is
relatively little variation between samples within the laccolith. Mock et al.
(2003) considered that these CSDs showed that these three minerals nucleated
and grew continuously during magma ascent and emplacement; in situ crystallisation was not detected as the laccolith cooled fast after emplacement at such
high levels. Minor changes in the textural parameters suggested that these were
two episodes of magma injection into the laccolith.
123
3.4 Typical applications
3.4.4.2 K-feldspar in granodiorite
Euhedral potassium-feldspar megacrysts up to 20 cm long are a striking and
relatively common component of many granitoid plutons, but their origin is
still controversial (see the review by Vernon, 1986). Many geologists hold that
the megacrysts are igneous in origin, that is, they grew from a granitic magma
and are phenocrysts. However, a minority view is that megacrysts are postmagmatic and grew from a circulating water-dominated fluid. Megacrysts are
fundamentally a textural phenomenon; hence lend themselves to CSD studies.
CSDs of megacrysts in the Cathedral Peak granodiorite (CPG) were studied by
Higgins (Figure 3.54a; 1999).
The CPG is the most voluminous part of the concentrically zoned Tuolumne
Intrusive Suite, in the Sierra Nevada of California (Bateman & Chappell,
1979). This suite started with the emplacement of dioritic magma. Each of
the following three phases was more felsic and was emplaced into the stillmushy core of the intrusion. An important component of the Cathedral Peak
granodiorite, adjacent parts of the Half-Dome granodiorite and the Johnson
Granite Porphyry is microcline megacrysts. The megacrysts are distributed
(a) K-feldspar in the Cathedral
Peak granodiorite
(b) K-feldspar CSDs
ln (population density) (mm–4)
–5
Groundmass
–10
–15
Megacrysts
–20
0
20 mm
20
40
60
size (mm)
80
Figure 3.54 (a) K-feldspar in a slab from the Cathedral Peak granodiorite.
The sample was stained, scanned and filtered for K-feldspar. (b) CSDs of
microcline crystals (Higgins, 1999). Megacrysts from five outcrops were
measured in the field with a ruler. Dimensions of groundmass microcline
crystals in the slabs were determined by automatic image analysis. Open
symbols are for areas with diffuse megacrysts whereas solid symbols are for
sites with ‘nests’ of megacrysts.
124
Grain and crystal sizes
heterogeneously with ‘nests’ and elongated areas up to a metre wide. The linear
zones meander for tens of metres across the outcrop and there does not appear
to be any contrast between vertical and horizontal sections. In general, threedimensional sheets have linear intersections with surfaces whereas string-like
forms have point intersections. Therefore, the surface distributions of the
megacrysts suggest that the volumes rich in megacrysts have sheet-and
string-like forms in three dimensions.
Somewhat surprisingly the CSDs of both the ‘nests’ of megacrysts and the
diffuse areas have similar shapes, but the nests have higher overall population
densities for all crystal sizes than the diffuse areas (Figure 3.54b). All CSDs
have a peak at 25 mm. The smallest megacryst that could be reliably measured
was 10 mm; hence the turn-down of the population density to the left of the
peak is not an artefact. If the CSDs were straight down to 10 mm then the
population density at 10 mm would have been 10 to 50 times higher, which is
readily distinguishable from the actual data. The limb to the right of the peak is
generally straight. The CSDs of the groundmass microcline are quite different
from those of the megacrysts. All three CSDs are quite straight, right down to
the smallest crystal that could be measured (1 mm) and much steeper than
those of the megacrysts. One slab was sufficiently large that the abundance of
smaller megacrysts could be determined, although the small numbers of large
crystals did not give a high precision (Figure 3.54a). The compatibility of the
two methods is confirmed by the overlap of the CSDs.
3.4.5 Other undeformed minerals
3.4.5.1 Diamonds
The value of diamonds in a deposit is critically dependent on the distributions
of sizes and quality of the crystals. Although the size distributions of diamonds
are commonly measured this information is usually secret. Of course, it is not
always clear if diamonds are phenocrysts or xenocrysts. However, their CSD
should tell us something about their regions of origin, the mantle. In this
environment one would expect that diamonds are coarsened, unless protected
inside large silicate grains. Hence, the relationship between microdiamonds
and macrodiamonds should be very interesting.
Measurement of diamond CSDs presents a number of unusual problems.
Diamonds are very rare in most rocks; hence a large sample size is needed.
Even then it is difficult to make sure that the sample is from a geologically
homogeneous unit – mixed diamond populations in the sample are almost
inevitable. In some deposits many diamond crystals are broken during emplacement, and it must be clearly established how broken crystals will be treated.
3.4 Typical applications
125
Commercially diamonds are separated by mechanical methods. Such a
process can lead to two problems – yield variations and broken crystals.
Although the yield of larger crystals is high, it may fall to zero for smaller,
commercially unimportant diamonds (<1 mm). If the yield variations are
known then compensation is possible, except where the yield is zero. As has
been shown earlier, it is important to establish the lower size limit of detection.
Upper size is also limited by the crushing process, which may break larger
crystals. However, in some deposits natural forces have already broken many
of the crystals during emplacement and these cannot always be distinguished
from breakage during processing.
Exploration studies commonly use total solution methods, in addition to
mechanical separation. A smaller sample is coarsely crushed and fused with
alkali carbonates. A number of minerals remain and the diamonds are selected
by hand from this concentrate. There is no lower limit to the size of diamonds
that can be recovered. However, the smaller sample size of solution methods
limits the largest crystals that can be recovered, as does the initial crushing of
the sample.
The most extensive study of diamond sizes is that of Rombouts (1995). He
considered that diamond sizes in most deposits follow a lognormal distribution. However, he pointed out that this commonly breaks down at extreme
sizes. He notes that many diamond deposits have exponentially increasing
crystal numbers versus diameter. This is what is observed in many other rocks
(a ‘straight’ CSD). Unfortunately, there is little published data that can be used
to verify this.
One study has been published on the sizes of diamonds in a lamproite from
Australia (Figure 3.55; Deakin & Boxer, 1989). Diamonds were extracted in
two ways: mechanical and solution. Few crystals were broken; hence a CSD
can be constructed. The two curves can be fitted together, if allowance is made
for the low yield of small stones in the mechanical separation. The resulting
CSD is relatively straight from 0.15–3.00 mm, but with a slight turn-up for
small crystals. This suggests that diamonds grew under conditions of increasing undercooling. Some nucleation and growth must have occurred just before
eruption. There is no evidence for coarsening, but it may have been concealed
by the last stage of growth.
Some eclogite xenoliths in kimberlites are very rich in diamonds (up to
0.5%). Several of these xenoliths have been studied by X-ray tomography
(e.g. Taylor, 2000). However, no information has been published on the
size distribution of diamonds in these rocks. It is true that the crystal
numbers are low; typically less than 30, but even so there would be interest
in such data.
126
Grain and crystal sizes
–6
Diamonds in lamproite:
Argyle, Australia
ln (population density (mm–4)
–8
Solution
separations
–10
–12
–14
Mechanical
separations
–16
–18
0
1
2
3
size (mm)
Figure 3.55 Diamonds in lamproite. Smaller diamonds were extracted
by solution methods, where mechanical separations were used for larger
diamonds. After Deakin and Boxer (1989).
3.4.5.2 Carbonates and chromites
Peterson (1990) studied the CSDs of several unusual minerals from carbonatite and other lavas of the 1963 eruption of Oldoinyo Lengai. He examined
the alkali carbonate minerals gregoryite ((Na2,K2,Ca)CO3) and nyererite
(Na2Ca(CO3)2) from carbonatite lavas and combeite (Na2Ca2Si3O9) from a
nephelinite lava. In the original paper Peterson did not describe how he
calculated the CSDs from the intersection data. However, in a later paper
part of the earlier data was recalculated (Peterson, 1996). His revised data have
straight CSD, right down to very small crystals. This is consistent with continuous nucleation and growth.
Waters and Boudreau (1996) examined chromite cumulates in the Stillwater
complex. The chromites occur in centimetric layers. Even in thin section the
crystals in the centre of the layer are clearly larger and more abundant. The
CSDs confirm that the centre has been coarsened, compared to the edges. This
behaviour is consistent with the DeHoff (1984) model of coarsening, as the
growth rate will be higher where the crystal edges are closer together. This type
of positive feedback, reinforcing early, weak structures has been explored by
Boudreau (1987, and subsequent papers) and may account for the well-known
‘inch-scale’ layering in the Stillwater complex.
3.4 Typical applications
127
3.4.5.3 Chondrules and minerals in meteorites
Study of meteorites differs from most other aspects of geology as data from
a small number of small samples are extrapolated to reveal the essential
character of large bodies. So far textural studies have been concerned with
three aspects of meteorite petrology: size distributions of chondrules in chondritic meteorites, CSDs of minerals in igneous-textured meteorites and CSDs
of magnetic mineral grains possibly of biological origin.
Most silicate-dominated meteorites contain at least some chondrules and
they may dominate the rock. Chondrules are spherical globules of silicate
melts (0.15–1 mm) that are present in almost all silicate meteorites. They
were formed by flash heating of dust in the solar nebula and their size
distribution has been used to infer the dynamics of the nebula (Rubin, 2000).
Given the importance of chondrules, there have been a number of studies of
their size distribution. In several early studies meteorites were disaggregated
and the true size of the chondrules measured. However, not many meteorites
are amenable to this treatment and information on the internal structure of the
chondrule is not available. Many more studies have determined chondrule
sizes from examination of sections. In many cases intersection data were not
converted to 3-D values but examined directly (e.g. Rubin & Grossman, 1987).
This is useful for showing differences in sizes between chondrules in different
meteor classes, but conclusions based on fits of the size distribution to wellknown models are weak. In other cases conversion equations derived empirically for sedimentary rocks were used (Friedman, 1958) that suggested that
chondrules were actual smaller than their intersections, which is clearly wrong.
Eisenhour (1996) proposed the use of a Saltikov-type method, modified to
account for the thickness of the thin section. He found that chondrule size
distributions that were formerly considered to be lognormal actually fitted
more closely the Weibull distribution. This is important as lognormal distributions have sizes down to zero, whereas Weibull distributions can have a
minimum size.
Shergottites are rare meteorites that are essentially basaltic rocks from
Mars. Many shergottites have olivine, pigeonite and/or augite phenocrysts
set in a fine-grained matrix dominated by plagioclase. CSDs of some of these
minerals have been examined in four different studies (Lentz & McSween,
2000, Taylor et al., 2002, Goodrich, 2003, Greshake et al., 2004). In all these
studies the CSDs were determined from intersection dimensions in thin sections and, where specified, converted using the inaccurate ‘Wager’ method.
The lower size limit was not specified, but appears to be close to the thickness
of a thin section, 0.03 mm. They found turn-downs for small crystals in some
128
Grain and crystal sizes
samples that may just be an artefact of this lower measurement limit. In
addition, they used equal sized intervals, hence many larger size bins were
empty. They then went on to connect up these isolated bins and concluded that
some of the CSDs had two different slopes. While this may be correct, the data
as presented in the papers do not prove this. Clearly this is an interesting
subject for CSDs studies, but perhaps the CSDs of many shergottites should be
studied together using more modern conversion and display methods.
There has been considerable interest in the possibility that some meteorites
may contain fossil bacteria from Mars, hence the first signs of extraterrestrial
life. One possible indicator of the presence of fossil bacteria is magnetite
crystals which are used by bacteria on Earth to orientate themselves (McKay
et al., 1996). The bacterial magnetite crystals have distinctive shapes and size
distributions, which could be used to identify Martian magnetofossils.
Thomas-Keprta et al. (2000) separated magnetite crystals from carbonate
globules in the Martian meteorite ALH84001 and examined them using a
transmission electron microscope. They also extracted magnetite crystals
from a bacterium for comparison. The magnetite crystals were sorted into
different shapes and the size distributions determined for these projected
images. Magnetite crystals in the bacterium were prismatic with a narrow
range in size and shape. Thomas-Keprta et al. (2000) considered that the
prismatic fraction of the meteoritic magnetite crystals also had a similar
range in shape and size which suggested that they may also be of bacterial
origin. There have been many criticisms of this conclusion, summarised by
Treiman (2003). One concerns the size distribution of the meteoritic magnetite
crystals: the size distribution appears to be much wider than that of the
bacterial magnetite crystals. Treiman (2003) suggested that the magnetite
formed by shock and thermal metamorphism of the carbonate. This could
form the observed lognormal distribution.
3.4.5.4 Hematite in pseudotachylytes
Pseudotachylytes are melts that are produced along faults during movements:
they both melt and solidify very rapidly. Petrik et al. (2003) have examined the
CSDs of hematite that crystallised from pseudotachylyte melts. The authors
converted intersection data in two different ways in different parts of the
paper: where stereologically correct methods are used the CSDs are straight
down to very small crystal sizes. However, elsewhere the authors used the
‘Wager’ conversion and interpreted the observed turn-down for small crystals
as indicating coarsening or closed-system crystallisation. It is more likely that
this feature is an artefact caused by both incomplete counting of small crystals
and incorrect conversion of data. The hematite CSDs indicate crystallisation
3.4 Typical applications
129
times of 10 seconds at the tips of the veins and up to 90 seconds within the
wider parts of the veins. Growth rates of hematite are about five orders of
magnitude faster than that of ilmenite in a lava lake.
3.4.5.5 Experimental studies
Although there has been much work in experimental petrology on the nature
of phases under different pressure and temperature conditions, there have been
few experimental studies that use the distribution of crystal sizes. Cabane et al.
(2001) examined the kinetics of coarsening (Ostwald ripening) of quartz in
simplified synthetic silicate melts. They did observe coarsening, but the kinetic
factors did not match that predicted by LSW theory (Lifshitz & Slyozov,
1961). They suggested that this may be caused by an initial transient regime
or that coarsening is not limited by diffusion, but by surface nucleation.
However, it may also be related to the ideal nature of the experimental
materials, or that the LSW model is not the best to apply to geological
materials (DeHoff, 1991, Higgins, 1998). Their work suggests that coarsening
may be very important just after nucleation, but that it is not important for
quartz crystals larger than 1 mm during geologically reasonable periods of
time. However, they admit that it may be a different story for other minerals
and magmas.
3.4.6 Undeformed metamorphic minerals
3.4.6.1 Garnet
Garnet is by far the most popular mineral for CSD studies in metamorphic
rocks, with the first study in 1963 by Galwey and Jones (1963). This may reflect
the ease with which it can be separated from other minerals. Most studies have
been on rock sections: the equant shape of most crystals minimises the stereological cut-section effect, but the intersection-probability effect must still be
considered (see Section 3.3.4.3). There have also been a number of studies in
three dimensions: these enable the spatial relations between grains to be
examined in detail. However, 3-D analysis is not always better than sectional
analysis as resolution problems with some X-ray methods may limit the size
range of crystals that can be examined.
Early studies established that garnets generally have a unimodal concavedown size distribution (Galwey & Jones, 1963, Galwey & Jones, 1966, Kretz,
1966a), although Kretz (1966a) did find one bimodal distribution. The
precise shapes of the CSDs are not easy to establish from these studies, but
appear to be between normal (Kretz, 1993) and lognormal (Cashman & Ferry,
130
Grain and crystal sizes
1988) by population density (¼ frequency, if the size intervals are constant).
X-ray tomographic studies have improved the quality of the CSDs by the
elimination of stereological effects (Carlson et al., 1995). In all studies there
appear to be no small crystals; however, the lower size limit of detection is
not generally well defined and hence this feature may be in some cases an
artefact.
There has been much debate on the interpretation of garnet CSDs. The
most popular theory is that garnet nucleated, grew and then became coarsened by Ostwald ripening following the LSW model (Cashman & Ferry,
1988, Miyazaki, 1996). Coarsening requires transport of material between
crystals and this is easy if there is a fluid present (silicate or water dominated),
or if the transport distance is short, as in monomineralic rocks. Carlson
(1999) has argued that Ostwald ripening is severely limited by diffusion
in polymineralic rocks under metamorphic conditions. He also points out
that the shape of the garnet CSDs is not that expected for LSW coarsening.
However, it has been pointed out that the LSW model may not be the
most appropriate for geological materials (DeHoff, 1991, Higgins, 1998). In
addition it is possible that deformation promotes circulation of fluids
by opening up channels. An interesting twist on the problem is suggested
by the data of Daniel and Spear (1999). They looked at Mn zoning patterns
in garnet crystals and concluded that they were formed by the nucleation
and growth of many crystals that subsequently rotated slightly and
coalescenced.
3.4.6.2 Accessory minerals
Accessory minerals are very important for geochemical modelling and geochronology, but little quantitative work has been done on their textures, except
for zircon (see Section 3.4.3.3). Zeh (2004) examined the CSDs of apatite,
allanite and titanite in four gneiss samples that were metamorphosed to maximum temperatures of 550 8C to 680 8C. He used a variety of 3-D techniques to
determine the CSDs of these minerals over a large size range. All CSDs are
unimodal, concave down. On a classical CSD diagram the peak shifts to larger
sizes as the slope of the right side fans around to shallower slopes. Zeh (2004)
constructed a complicated growth history using the growth model of Eberl
et al. (1998) followed by coarsening. However, the CSD patterns closely
resemble the progression seen in other rocks that has been explained just
by coarsening (e.g. Cashman & Ferry, 1988, Higgins, 1998), especially that
following the ‘Communicating Neighbours’ equations of DeHoff (1991).
Coarsening in metamorphic rocks may be ubiquitous; hence the nature of
the original CSD is not observable.
3.4 Typical applications
131
3.4.6.3 Crystals in migmatites
Migmatites are rocks produced during partial melting. There are a number of
aspects of interest here: the size-dependence of melting can be investigated, but
is usually concealed by later processes. Berger and Roselle (2001) instead
examined the CSD of cordierite produced by melt-producing reactions. They
found that it was straight and proposed a nucleation and growth model similar
to that for crystallisation of minerals from magma. Plagioclase in a leucosome
had a concave down CSD suggesting coarsening. K-feldspar had a bimodal
CSD that was not easy to explain. Berger and Roselle (2001) only examined
three samples in all: clearly this material has a lot of promise for further study.
3.4.6.4 Hydrothermally grown crystals
Study of crystal sizes in hydrothermal crystallisation experiments by Eberl
et al. (1998, 2002) led to their proposal that all CSDs are lognormal and that
this necessitates a new growth theory, which they entitled ‘the law of proportionate effect’. However, as has been seen above, there are many processes that
can affect CSDs and it seems simplistic to force fit one model, especially when
it has no experimental basis. In addition, some of the studies of Eberl and his
coworkers have been hampered by incorrect conversion equations from intersection measurements to CSDs. Finally, the frequency diagram used in these
studies is not very sensitive to variations in CSD shape. Bindeman (2003) has
shown that some CSDs are indeed lognormal, but this is more likely produced
by coarsening or repeated cycles of growth and solution (see Section 3.4.3.2).
3.4.7 Deformed and broken crystals and blocks
3.4.7.1 Olivine in mantle lherzolites
Mantle samples fall between the simple classification of igneous and metamorphic rocks – they are clearly igneous in origin, but have been deformed.
There are few studies of the CSDs of minerals in mantle samples despite the
volumetric dominance of the mantle in the Earth, the enormous number of
fabric studies and the need for size data in geophysical modelling. Armienti
and Tarquini (2002) examined the CSDs of olivine in suites of spinel and
garnet lherzolites with a range of textures. They found that the CSDs followed
a power law distribution over two orders of magnitude of size. Granoblastic
textured samples had fractal dimensions around 2.4, which compares with
values of 2.6 for fragmentation models (see Section 3.3.7.1). Porphyroclastic
and cataclastic textured samples had fractal dimensions up to 3.8, but the
interpretation of this is not yet clear.
132
Grain and crystal sizes
3.4.7.2 Dynamic recrystallisation
The fabric of naturally deformed minerals has been studied extensively generally by using the mean grain size. Stipp and Tullis (2003) examined experimentally the relationship between mean grain size in quartzite and flow stress.
They ensured that deformation reached equilibrium and that stress was measured accurately, but unfortunately did not apply stereological corrections to
their grain sizes values. They found that mean grain size was well correlated
with flow stress within the dislocation creep regimes. A slightly different
correlation was observed at greater flow stresses.
Herwegh et al. (2005) have pointed out that the full grain size distribution is
important in controlling creep behaviour, although it is rarely measured. In
mylonites deformation of the rock reduces grain size at the same time as
recrystallisation repairs the damage. The result is an equilibrium grain size
distribution that depends on the temperature and the strain rate. Herwegh
et al. (2005) examined the CSDs of calcite from mylonites in the Helvetic Alps.
They found concave down frequency distributions, which they summarised
using mean, mode, skewness and kurtosis. They showed that deformation
follows grain-size-sensitive mechanisms, such as volume or grain boundary
diffusion and grain boundary sliding.
It seems logical that the presence of water should increase the growth rate of
minerals during deformation and recrystallisation. Experimental studies of
olivine proved this to be true (Jung & Karato, 2001). The size of olivine
subgrains was determined from their intersection lengths. These values were
multiplied by 1.5 to estimate the 3-D size, but no correction was made for the
intersection-probability effect. These modified intersection lengths had a lognormal distribution. Olivine deformed under dry conditions had a mean
intersection size of 3 mm, whereas in the presence of water the mean intersection size was 15 mm.
3.4.7.3 Broken blocks and crystals
Rocks are fragmented during engineering and mining operations: hence, the size
distribution of fragments has economic implications, such as in the processing of
mining tailings. Turcotte (1992) has compiled a number of studies of the size
distribution of broken rock (Figure 3.56). The data covered an unusually wide
range of sizes – up to four orders of magnitude. He concluded that explosively
crushed rock had a fractal dimension of about 2.5. This compares closely to a
simple fragmentation model that produced a fractal dimension of 2.58.
A fractal model of fracturing has not been universally accepted: Suteanu
et al. (2000) mention that some authors have considered that fracturing
133
log (number of fragments < size)
3.4 Typical applications
0.01
Coal
Granite
Basalt
D = 2.50
D = 2.50
D = 2.56
0.1
1
10
log (size) (mm)
100
1000
Figure 3.56 Fragment sizes of artificially broken rock (after Turcotte,
1992). Size is cube root of fragment volume. D is the fractal dimension.
produces multimodal distributions. Their experiments suggest that this may
reflect the way data have been presented, however, they also offer the possibility that size distributions are multifractal – that is, the fractal dimension
changes with the size. Data quality must be very high to be able to prove
multifractal size distributions, but a change in fractal dimension is to be
expected around the typical grain size of the rock.
Engineering geologists commonly have to deal with heterogeneous rocks
that are comprised of hard blocks of variable size in a weaker matrix. Such
‘block-in-matrix rocks’ (bimrocks) would be described by geologists as conglomerates, breccias and fault gouges. However, an engineer is more interested
in the mechanical properties than in the origin of the rock. Hence, it is
important to be able to quantify the size distribution of the blocks in bimrocks.
Medley (2002) has shown that this can be done from measurements of blocks
intersected by drill holes or surface exposures using the same stereological
methods that are used for examining crystal sizes.
Kimberlites are magmatic rocks that are commonly loaded with xenoliths,
and their size distribution can be used to determine their mode of origin.
Kotov and Berendsen (2002) examined smaller xenoliths in a kimberlite pipe
from Kansas. They present few details of their results but conclude that
the sizes of xenoliths follow a fractal distribution, here transformed to the
Pareto distribution so that it could be tested statistically. This distribution is
134
Grain and crystal sizes
consistent with a model for the origin of the xenoliths by deep explosion
without further breakage during transport. In half of the samples minor
divergence from this distribution could be ascribed to Bagnold effect sorting.
Fault gouge is the powdery rock developed along faults during earthquakes.
Measurement of the particle size distribution (PSD) of gouges is not easy as
they are very fine grained and particles tend to aggregate. Wilson et al. (2005)
used a laser particle-size analyser to determine PSD and found that the PSD
varied over time as the aggregates broke down. However, at no time was the
distribution fractal as has been noted for larger size fractions. They concluded
that fault gouge did not form by wear and attrition of the fault surfaces, but by
dynamic rock pulverisation during each earthquake. In addition, creation of
the large surface area of the gouge particles could easily take a sizeable
proportion of the energy of an earthquake.
Some crystals in volcanic rocks are broken (Allen & McPhie, 2003). So far
study has been limited to the overall abundance of broken crystals, rather than
their size distribution. In one ignimbrite 67%–95% of crystals were broken,
but in intermediate and silicic lavas and tuffs a more typical value is 5%–10%
(Allen & McPhie, 2003).
Notes
1. There has been debate in petrology about the use of the terms solution and melting. We insist
that children are incorrect when they talk of sugar melting in water, but we commonly talk of
granite melting to a silicate liquid. To a chemist a liquid must be a pure substance; hence in
chemical terminology, both these examples are dissolution and silicate liquids are in fact
high-temperature solutions. Hence, rocks do not melt, they dissolve. However, we all know
what we mean by melt, even if we use different terms.
2. There is another terminology problem here. To a physicist a fluid is just a liquid, gas or supercritical fluid. However, many geologists use the term fluid if the substance is dominated by
water, and melt if it is dominated by silicates (or carbonates, sulphides, etc.). In actual fact
there is a continuum of fluids with compositions from pure water to pure silicate. And, as
noted above, they are all actually solutions.
3. Also called the Schwartz–Saltykov method.
4
Grain shape
4.1 Introduction
Grain shape is something that is easy to express qualitatively, but difficult to
quantify precisely (Costa & Cesar, 2001, Verrecchia, 2003). In earth sciences
we do not generally need extremely precise measures of shape, because the
objects we deal with are commonly not perfectly regular. However, quantification of aspects of shape can help in understanding the petrogenesis of rocks.
The subject naturally falls into two domains. For crystalline rocks (igneous
and metamorphic rocks, chemical sediments and hydrothermal ore deposits)
we are concerned with the shape or habit1 of crystals, whereas in clastic rocks
we want to quantify the shape of clasts and grains. The methodology of these
two fields is partly shared but could benefit from more exchange.
The overall shape of crystals reflects growth, solution and deformation.
Crystals that grew unimpeded from a fluid may have many different forms,
but all are ultimately controlled by the crystal structure. We usually think that
faces bounding such grains should be flat, but in some minerals growth faces
are curved. In thin section crystal outlines are qualitatively classified as euhedral, subhedral or anhedral.
The habit of euhedral crystals has been studied for a long time (see the
introduction in Sunagawa, 1987a): in 1669 Nicolaus Steno proposed that the
interfacial angles of a crystal were constant and that the external form of a
crystal growing in a fluid depends on the relative growth rates of the different
faces. In many mineralogical and petrological studies the habit of crystals
has its own interest (e.g. Sunagawa, 1987b). For instance, it is commonly
observed that crystal habit varies with growth environment and size (see the
review in Kostov & Kostov, 1999): plagioclase crystals in the groundmass
of lava are tabular (commonly misnamed ‘lath-shaped’) whereas those in
more slowly cooled rocks are more equant. Rapid growth forms, such as
135
136
Grain shape
spherulites, dendrites and swallowtails, are not treated here as they are difficult
to quantify.
Crystal shapes can help in the quantification of other textural parameters:
an estimate of shape is needed to carry out some types of stereological corrections (e.g. Higgins, 2000). For this purpose overall crystal shapes are simplified
and defined in terms of the aspect ratio and degree of rounding. This simplified
scheme can also be used to examine crystal shape variations between different
petrologic environments.
Grain boundaries and dihedral angles are other quantifiable aspects of
crystal shape: grain boundaries may be flat, curved or sutured, and all these
can be quantified. Dihedral angles are the angles at the corners of crystals
(Smith, 1948). It has been shown that the values of such angles are a measure
of the textural maturity of a rock (Elliott et al., 1997).
4.2 Brief review of theory
4.2.1 Shapes of isolated undeformed crystals
4.2.1.1 Equilibrium forms
Crystal growth and final habit are controlled by many factors (Baronnet, 1984,
Sunagawa, 1987a): a good point of departure is the equilibrium form of the
crystal. This form is purely theoretical and is that in which the surface energy
(or surface tension) of the crystal is minimised (Wulff, 1901, Herring, 1951a).
Unlike a liquid, the surface energy of a crystal varies with orientation; hence
the equilibrium form of a crystal is never a sphere. The energy associated with
a crystal face depends on the density of lattice points – the reticular density.
Lattice points are not necessarily atoms, but points in space that have the
identical environment in the same orientation. A crystal face is more developed
if it has a high reticular density; that is it intersects a large number of lattice
points (Bravais law). A more recent theory uses the concept of periodic bond
chains (PBC; Baronnet, 1984). A PBC is an uninterrupted chain of strong
bonds between atoms. On this basis faces can be classified: F (flat) faces have
two or more PBCs, S (stepped) faces have one PBC, and K (kinked) faces have
no PBCs at all (Figure 4.1). It may seem counter-intuitive, but the largest faces
of a crystal have the lowest growth rates. The F faces grow by layer mechanisms and hence grow at the slowest rates. Therefore, the facets of a crystal are
generally the F faces.
The equilibrium form of a crystal is controlled by the crystalline structure
and hence does not depend on the environment in which the crystal grew.
However, it does depend on extensive variables, such as temperature. At
137
4.2 Brief review of theory
F
F
S
F
F
F
S
K
S
F
Figure 4.1 Periodic bond chains and crystal structure. F ¼ flat faces;
S ¼ stepped faces; K ¼ kinked faces.
higher temperatures the energy difference between different faces becomes less
and it becomes energetically feasible to have curved faces. At the extreme this
tends to a sphere with minor facets. The equilibrium forms of common
minerals broadly resemble the observed forms in the expression of faces, but
the importance (relative size) of the faces does not always agree.
4.2.1.2 Growth forms
There are many different growth forms of crystals, even under identical
thermodynamic conditions (Sunagawa, 1987a). The broad classification of
crystals whose growth has not been impeded by the presence of other materials
is into spherulitic, dendritic, hopper and polygonal. The different faces of
polygonal crystals may be developed to different extents, giving different
polygonal habits. Where crystals are growing in the presence of a second
phase they may be incorporated into the second phase as chadocrysts, or
they may stop growth, leading to anhedral (irregular) forms. Some of the
latter may have curved faces formed by the simultaneous growth of two
crystals.
The growth habit of crystals is fundamentally controlled by growth rates of
the various faces or directions. Such growth rates are controlled by intensive
parameters such as temperature, pressure and chemical potential gradient
(Figure 4.2; Kouchi et al., 1986, Kostov & Kostov, 1999). Where crystals are
surrounded by other crystals, for instance in the sub-solidus, other parameters
are important, most significantly the interfacial energy (see dihedral angles
below; Hunter, 1987).
4.2.1.3 Solution forms
The solution forms of crystals have not been studied in as much detail as
growth forms. A common observation of solution forms is within crystals, as
revealed by zoning. For instance, the optical properties of plagioclase make
138
Grain shape
o
growth rate
a
Ratea
Rateo
growth parameters
Figure 4.2 A schematic example of the development of crystal forms related
to different growth rates under different growth parameters. The a and o faces
have growth rates that vary in a different way to growth driving force
parameters, such as undercooling (supersaturation), chemical environment
or composition.
it easy to observe minor variations in composition. Resorption surfaces that cut
across fine composition zones are commonly encountered, especially in volcanic rocks (e.g. Stamatelopoulou-Seymour et al., 1990). In all cases the solution
form is rounded. That is, the radius of the corner is increased progressively by
the consumption of the faces. The solution surface is frequently rough, reflecting the effects of dislocations and impurities in the crystal. Other minerals,
such as quartz, pyroxene and olivine have resorbtion surface zones similar to
those of plagioclase, but they are not so easily seen, except with techniques like
Nomarski microscopy (see Section 2.5.3.3). Most natural diamond crystals
have suffered solution during transport from the mantle, which is expressed
by curved crystal surfaces. The process starts with loss of the sharpness of the
edges and proceeds to the faces, which become curved (Sunagawa, 1987b).
Extensive solution leads to rounded crystals.
4.2.2 Shapes of deformed crystals
The shape of deformed crystals, like their size, is a balance between the
processes that deform crystals (see Section 3.2.5), which give the crystal an
excess of energy, and the processes that will dissipate that energy by the
elimination of lattice defects (relaxation) and the reduction of surface area
139
4.2 Brief review of theory
(a)
Crystal 2
(b)
(c)
Crystal 1
Figure 4.3 Schematic development of sutured grain boundaries between
two crystals. (a) Initial state after deformation. Circles are lattice defects.
(b) Bulging of the crystal boundary removes a defect from the adjoining
crystal. The new boundary is then close enough to capture other defects.
(c) The process will continue while it is energetically favourable to make more
grain surface and lose lattice defects.
per unit volume. In deformed rocks the energy associated with lattice defects
is typically two to four orders of magnitude greater than that associated with
excess surface area (e.g. Duyster & Stockhert, 2001). During uniform deformation equilibrium may be reached between these processes, giving a constant
crystal shape that depends on the deformation style and rate, and the temperature and fluid content. Of course, such information may be masked by
subsequent deformation under different conditions. Deformation mechanisms
have been discussed briefly in Section 3.2.5. Recovery of surface energy by
coarsening (ripening) has also been discussed with reference to crystal size
(see Section 3.2.4). However, the role of lattice defects has yet to be discussed.
Lattice defects have excess energy compared to other parts of the crystal
lattice, hence will be removed during relaxation. Recovery of such lattice
defects close to the edge of a crystal can occur by grain boundary migration
(Kruhl & Nega, 1996). A suitably located defect can be eliminated by removal
of that part of the grain and reconstitution of the atoms into the adjacent grain
(Figure 4.3). The overall effect is a migration of the grain boundary and the
bulging of one grain into another, leading to a sutured grain boundary. The
energy expended in creating a larger grain surface area is less than the energy
recovered from the elimination of the defect.
4.2.3 Dihedral angles of crystals in rocks
So far the shape of isolated crystals has been discussed, but in most rocks
crystals touch. If such a material is completely solid then four crystals meet at
most crystal corners and three meet along most crystal edges. In the plane
normal to the edge the angle of the traces of the crystal boundaries is the
dihedral angle (DA).
140
Grain shape
Crystals will grow unimpeded in magma or liquid until they impinge
another crystal (Figure 4.4). The impingement DAs for such textures are
very variable but the population of DAs has a mean of 608 (Figure 4.5). This
texture is regarded as the departure point for further textural development in
solidifying systems (Holness et al., 2005). Such textures should be developed in
simple magmatic systems without magma flow or imposed stress. They may
also develop in some hydrothermal ore deposits and chemical sediments.
B
C
A
B
A
Liq
C
Figure 4.4 Impingement angles in growth textures. Here, three different
phases (A, B and C) have grown from a liquid without equilibration and
a portion of the liquid remains (Liq). Some of the many dihedral angles
associated with each phase are shown by arcs. In a simple growth texture
like this these dihedral angles are referred to as impingement angles.
40
Su
b-s
oli
du
se
qu
es
en
t
eq
u
standard deviation
ilib
ra
tio
n
Impingement
(growth textures)
ilib
rat
i
M
el
t-p
r
on
Melt-present equilibrium
Solid-state equilibrium
0
0
60
median angle (°)
120
Figure 4.5 Changes of the median and standard deviation of dihedral angles
from initial impingement (growth) textures in response to melt-present and
sub-solidus equilibration (from Holness et al., 2005).
141
4.2 Brief review of theory
At equilibrium the DA reflects the difference in surface energy (interfacial
tension) between the phases. If three phases are isotropic and meet at an edge,
then (Smith, 1948, Smith, 1964)
g1
g
g
¼ 2 ¼ 3
sin a1 sin a2 sin a3
where g1, etc. are the surface energies between the phases along a grain
boundary, and a1, etc. are the DA between the phases on the opposite side of
the edge (Figure 4.6a). Liquids are isotropic, but all minerals are physically
anisotropic, hence the surface energy varies with direction and even at equilibrium there will be a range of DAs (Herring, 1951b, Laporte & Provost, 2000).
Experimental work suggests that the degree of anisotropy broadly correlates
with the overall symmetry of the minerals: it is lowest in isometric minerals like
spinel and highest in monoclinic minerals like diopside (Laporte & Watson,
1995, Schafer & Foley, 2002). Some studies have shown that true F-faces
(Figure 4.1) may form where one of the phases is a fluid, leading to further
complexity in the anisotropy of surface energy (Cmiral et al., 1998, Kruhl &
Peternell, 2002) although this may not always be the case (Hiraga et al., 2001).
In a polymineralic rock many different mineral combinations are possible,
each with their own range of equilibrium DAs (Figures 4.6b, c, d).
(a)
(b)
γ2
γ1
X
(c)
α3 α1 γ
3
α2
A
A
Y
120°
A(A–A )
A
B
A
Z
130°
A(A–B )
A
(d)
A
100°
A(A–C )
A
Three other
combinations
are possible:
A(B–B )
A(C–C )
A(B–C )
C
Figure 4.6 Equilibrium dihedral angles in a three-phase system. (a) If three
different isotropic phases X, Y, Z, meet at an edge then the DAs a1, a2, a3 are
simply related to the interfacial energies of the grain boundaries g1, g2 and g3.
(b) If the rock is monomineralic then the mean equilibrium DA will be 1208.
The DAs will be variable because the surface energy of minerals is not
isotropic. (c, d) Here, the rock has three phases A, B and C, one of which
may be a liquid. In this case the equilibrium DA for each mineral varies with
the nature of its neighbours, as well as the anisotropy. In a three-phase rock,
six different combinations are possible, of which three are shown.
142
Grain shape
Clearly, surface energies are important in the development of texture and
their values should be determined. Such values may be calculated for crystals
in a vacuum, an uninteresting value for most geological studies. DAs can
provide useful estimates of the ratios of surface energies (Kretz, 1966b,
Vernon, 1968, Hiraga et al., 2002), but it is much more complicated to measure
the actual surface energies. Surveys in materials science suggest that values
are of the order of 1 Jm2 for many contacts between crystalline materials
(Sutton & Balluffi, 1996). The surface energy of olivine–olivine contacts was
determined from a mantle sample to have a maximum value of 1.4 Jm2
for the greatest degree of crystal misorientation (Duyster & Stockhert, 2001).
Surface energies between solids and liquids are much lower than these values,
because the liquid is structurally isotropic. Mungall and Su (2005) found
that the surface energy between sulphide and silicate liquids was 0.5–0.6 Jm2.
They note that the surface energy of silicate mineral–silicate liquid contacts
should be much lower.
Dihedral angles can be used as a measure of the degree of textural equilibrium in a rock (Elliott & Cheadle, 1997, Holness et al., 2005). A rock is in
textural equilibrium if ‘the surface topology of the grains is in mechanical and
thermodynamic equilibrium’ (Elliott et al., 1997). If a rock has coarsened in a
stress-free environment then it will have approached textural equilibrium (see
Section 3.2.4). The attainment of textural equilibrium can be assessed from
the distribution of grain sizes, from the curvature of the grains, or from the
distribution of DAs. In general DAs are the most sensitive to equilibration and
grain shape and size the last to respond (Laporte & Provost, 2000). In a rock
composed of only one phase the mean DA is 1208 (Figure 4.6b). If several
phases are in equilibrium then the DA for a particular mineral will depend on
its neighbours (Figures 4.6c, d) in addition to the anisotropy of surface energy
(Kretz, 1966b, Vernon, 1968, Elliott et al., 1997, Holness et al., 2005).
DAs are of particular interest where one of the phases is a liquid, which is
isotropic. In this case the DA is a measure of the wetability of the surface by the
liquid: if we consider again our ideal rock, composed of one isotropic mineral,
then if the DA between the solid phase and a liquid is less than 608 then the
liquid will form an interconnected network along three-phase junctions. This
means that the connectivity threshold of the melt is zero. A survey by Smith
(1964) suggests that this is the case for some minerals, aggregates of which are
permeable even at very low degrees of melting. Conversely, if the DA is greater
than 608 then the liquid will aggregate at the intersections of grains, and the
connectivity threshold is much higher.
Calculation of rates of textural equilibration or coarsening requires absolute
values of surface energies, which cannot be determined from DAs. However,
143
equilibrium melt fraction (vol%)
4.3 Methodology
30
Laporte & Watson
(1995) – bcc model
20
Bulau et al. (1979)
Beere (1975)
Park & Yoon (1985)
10
Jurewicz &
Watson (1985)
0
0
20
40
60
dihedral angle (°)
80
100
Figure 4.7 Minimum-energy melt fraction in partially molten rocks at
equilibrium against dihedral angle (modified from Beere, 1975, Bulau et al.,
1979, Jurewicz & Watson, 1985, Park & Yoon, 1985, Laporte & Watson,
1995, Ikeda et al., 2002).
approximate values estimated from an assessment of relevant experimental
studies (Sutton & Balluffi, 1996, Holness & Siklos, 2000), suggest that textural
equilibrium over a scale of millimetres could be achieved in partially molten
silicate rock over tens of years.
A number of experimental studies have suggested that at equilibrium there
is a minimum-energy melt fraction (MEMF) in a partially molten rock that
depends on the DA (Figure 4.7; Jurewicz & Watson, 1985, Ikeda et al., 2002).
If the actual amount of melt is less than the equilibrium melt fraction then melt
will be drawn into the crystal aggregate. However, if the material contains
more melt than the equilibrium fraction, then the melt will be expelled and
segregate outside the aggregate. These effects could cause the commonly
observed clustering of crystals.
4.3 Methodology
4.3.1 Three-dimensional analytical methods
Grain shape is a property of the surface of the grain. It can be examined
precisely using three-dimensional methods, if the grain surface can be separated perfectly from adjoining grains by physical or analytical means. This is
the best way to look at the development of faces in minerals, and complex
shapes such as dendrites. It is also necessary if there is a significant range in
144
Grain shape
grain shape within a sample, for example for different grain sizes. The following
3-D methods have been used:
*
*
*
*
*
The shape of individual grains can be determined by serial sectioning (see
Section 2.2.1).
In transparent materials, such as volcanic glasses, the shape of crystals can be
measured directly with a regular or confocal microscope (see Section 2.2.2).
X-ray tomography can be used to determine the shape of grains if adjoining grains
can be separated (see Section 2.2.3).
The shape of grains is preserved if they can be extracted intact from rocks by
mechanical disintegration or partial solution (see Sections 2.5 and 2.6). Natural
weathering can sometimes lead to the extraction of measurable grains (Higgins, 1999).
Rocks commonly fracture along zones of weakness, which may include crystal
surfaces or schistosity in clasts. Hence, aspects of the shape of individual grains can
be commonly measured from such special sections exposed on irregular fractured
blocks. This is particularly useful for tabular grains.
4.3.2 Section and surface analytical methods
It is not always possible to separate grains either chemically, mechanically or
analytically for 3-D analysis. In these cases two-dimensional intersection or
projection methods may give better results. Larger objects may be sectioned
individually along special directions: this is important for looking at internal
surfaces. However, generally the outlines of many grains are examined statistically and the mean shape determined. These methods may not be applicable
if a range of shapes is present.
The shape of grains can be determined from images acquired from rock
surfaces, slabs and sections, using a variety of techniques described in
Section 2.5. Such images are processed manually or automatically to produce
classified images (see Section 2.6). Finally, the classified images are processed
to extract grain outline parameters. If the section is examined using techniques
that can reveal the orientation of the crystal lattice, such as polarised light
microscopy or EBSD, then the relationship between the shape and the lattice
orientation can be determined.
Commonly data from many grains with different orientations are combined
and the results interpreted in terms of grain shape. However, some shape
parameters are difficult to measure by these methods and it may be better
to examine grains with special orientations that are parallel or normal to the
plane of the section. For crystals such orientations may be identified using
optical methods such as birefringence and twinning. A universal stage can also
be used to rotate crystals into the useful orientations (see Section 5.5.1).
4.3 Methodology
145
Changes in the shape of a crystal during its lifetime may be observed
from zoning patterns, or surfaces rich in inclusions inside the crystal
(Sempels, 1978, Vavra, 1993). Such patterns are best revealed by regular
optical methods, backscattered electron images (see Section 2.5.2.1), cathodoluminescence (see Section 2.5.2.3), Nomarski interference (DIC; see
Section 2.5.3.3) or laser inferometry methods (Pearce et al., 1987). Of course,
important solution events may remove earlier history, so the story may
be fragmentary. For quantitative studies of changes in morphology it may
be necessary to use sections with special orientations (Pearce et al., 1987,
Sturm, 2004).
4.3.3 Extraction of overall grain shape parameters
4.3.3.1 Overall axial ratios
Grains can have very complex shapes but for some purposes all that is needed,
or all that can be currently determined, are the overall shape in terms of its
aspect ratio. There are a number of different definitions (see Figure 3.26).
*
*
The dimensions of the smallest parallelepiped that encloses the grain. The aspect ratio
then has three parameters: the short, intermediate, and long dimensions (S : I : L).
The dimensions of a triaxial ellipsoid with the same moments of inertia as the grain.
The aspect ratio is then the ratio of the minor (short) : intermediate : major (long)
axes (S : I : L).
The aspect ratio of either model can be precisely determined from threedimensional data for each grain. The limitations of three-dimensional methods
have been described above; hence it may be necessary to use two-dimensional
data from intersections or projections. For practical reasons it is easier
to divide the total aspect ratio into the ratios S/I and I/L, both of which vary
from 0 to 1.
In sedimentology the SIL values are commonly reduced to a single parameter, the sphericity:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
sphericity ¼ ðIS=L2 Þ
Sphericity has a value of one for a sphere and decreases towards zero for more
elongate clasts.
4.3.3.2 Intersection data
If grain shapes are convex, constant in shape and simple then parts of the
overall aspect ratio may be estimated statistically from grain intersections
146
Grain shape
using the distribution of the ratio of intersection widths and lengths (w/l),
supplemented with other data where necessary. Obviously such methods
cannot be used for concave or branching crystal forms, which must be measured by 3-D methods such as X-ray tomography or serial sectioning (see
Section 2.2).
One simple method of estimating S : I : L is derived from a model of grains as
parallelepipeds (Higgins, 1994). For tablet-shaped grains with no preferred
orientation (massive) the mode of the w/l is equal to S/I (Figure 4.8a). For
prisms the results are not so easy to interpret: w/l ratios have a wide distribution with only a minor peak at S/L (Figure 4.8b). A model based on ellipsoids
(a) Tablets
(b) Cube and prisms
1:10:10
frequency
1:1:2
1:1:1
1:5:5
1:1:5
1:2:2
0
0.2
0.4
1:1:10
0.6
0.8
1.0 0
0.2
0.4
intersection width / intersection length
(c) Oblate ellipsoids
0.6
0.8
1.0
0.8
1.0
(d) Prolate ellipsoids
frequency
1:10:10
1:1:2
1:5:5
1:2:2
1:1:5
1:1:10
0
0.2
0.4
0.2
0.4
0.6
0.8
1.0 0
intersection minor axis / intersection major axis
0.6
Figure 4.8 (a, b) Distribution of intersection width/length for randomly
oriented planes intersecting a parallelepiped (after Higgins, 1994).
Intersection dimensions are those of the smallest rectangle that can be fitted
around the intersection. Representative crystal intersections are shown in
Figure 3.27. (c) Intersections of a randomly oriented plane with an oblate
ellipsoid (disc shaped; this study). (d) Intersections of a randomly oriented
plane with a prolate ellipsoid (rugby/American football shaped; this study).
147
4.3 Methodology
Table 4.1 Modal values of different intersection parameters for populations
of parallelepipeds with different fabrics (Higgins, 1994). S ¼ short dimension,
I ¼ intermediate dimension, L ¼ long dimension. For some orientations of strong
fabrics values are invariant: these are shown in bold italic type.
Fabric type and
section orientation
Mode of
intersection
length (l)
Mode of
intersection
width (w)
Mode of
intersection
width/length (w/l)
Massive
Foliated – normal
Foliated – parallel
Lineated – normal
Lineated – parallel
F and L – normal
F and L – parallel
I
I
L
I
L
I
L
S
S
I
S
I
S
I
S/I
S/I
I/L
S/I
I/L
S/I
I/L
rather than parallelepipeds has a broadly similar distribution of intersection
minor axis/intersection major axis ratios, equivalent to w/l. For oblate ellipsoids the ratio S/I (ellipsoid minor axis/ellipsoid intermediate axis) can be
determined from the mode of the intersection minor axis/intersection major
axis ratios (Figure 4.8c). As for prisms, the shape of prolate ellipsoids cannot
be determined from the distributions of the intersection minor axis/intersection major axis ratios (Figure 4.8d).
Rocks with strong fabrics may have different w/l modal values depending
on the orientation of the section with respect to the foliation or lineation
(Table 4.1).
If a sample has a strong foliation or lineation and a section parallel to
the fabric is available than the ratio I/L can be determined also. However,
more commonly the ratio I/L must be determined from the distribution of the
w/l ratios. Higgins (1994) showed that for parallelepipeds this ratio can be
estimated from the w/l distributions as follows:
I=L ¼ 0:5 þ ðmean w=l mode w=lÞ=standard deviation w=l
However, this equation does not give accurate results for near equant shapes.
Garrido et al. (2001) did a very similar modelling of parallelepipeds and
found a graphical relationship between the mean w/l, mode w/l and I/L. The
following equation was derived from their graph.
I=L ¼ ðmean w=l þ 1:09ðmode w=lÞ2 1:49ðmode w=lÞ þ 0:056Þ=0:126
148
Grain shape
Both methods need an accurate estimate of the mode w/l which is not easily
available for most measured CSDs. In many cases it is best to measure I/L
from crystals with special orientations (see Sections 4.3.1 and 4.3.2).
As was mentioned before the shape of prismatic crystals is not easily
determined from w/l distributions. The best way to determine the shape of
prismatic crystals is from the examination of crystals with special orientations:
that is, with the long axis parallel to the section.
Some stereological methods for calculating crystal size distributions
(e.g. CSDCorrections; Higgins, 2000) require also an estimate of the degree
of rounding of crystals. For the models described above this parameter just
modifies the shape model used for the conversion by combining intersections
from parallelepipeds and ellipsoids.
4.3.3.3 Projection data
In some situations grain outlines are seen in projection and not in section. For
instance, grains may be completely contained within a thin section. The projected outline width/length (w/l) can be used to determine aspects of the overall
aspect ratio, as for intersection data, but the relationships are different. The
parallelepiped model assumes that blocks are randomly oriented. Tablets have
broad distributions of projected w/l that cannot be used to determine the
aspect ratio of the grains (Figure 4.9a). However, prisms have well defined
w/l peaks for which the mode ¼ I/L (Figure 4.9b). For shapes between tablets
and prisms the S/I ratio is not easily accessible and is best determined from
grains with special orientations.
A model based on randomly oriented ellipsoid grains gives similar, but not
identical results to the parallelepiped model: for oblate ellipsoids the modal
projected minor axis/projected major axis has a peak at I/S, but it is narrow,
with few intersections (Figure 4.9c). Hence, in practice it is not possible to
determine the value of I/S for such shapes. However, for prolate ellipsoids the
peak at I/L is well defined and the mode is easily measurable in real situations
(Figure 4.9d). The I/S ratio of triaxial ellipsoids is not easily accessible from
projection data.
If grains are deposited as a layer just one grain deep then their outlines
can be easily imaged and measured. Here, the grains are not randomly
oriented, but lie with the I–L plane horizontal. In this situation, the value
of the w/l will be exactly equal to I/L, however S/I is not accessible from
the image. If all grains are assumed to have the same composition then the
total mass of the sample can be used to estimate the value of S/I (Mora &
Kwan, 2000).
149
4.3 Methodology
(a) Projected tablets
1:1:10
(b) Projected prisms
frequency
1:1:5
1:1:2
1:2:2
1:10:10
0
1:5:5
0.2
0.4
1:1:1
0.6
0.2
0.8
1.0 0
projected width / projected length
1:1:10
(c) Projected
oblate ellipsoids
0.4
0.8
1.0
0.8
1.0
(d) Projected
prolate ellipsoids
1:2:2
frequency
0.6
1:1:2
1:1:5
1:10:10
1:5:5
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
projected minor axis / projected major axis
Figure 4.9 (a, b) Parallelepiped model of randomly oriented grains in
projection (this study). The projected length and width are measured as the
dimensions of the smallest rectangle that can be placed around the outline.
(c, d) Ellipsoid model of randomly oriented grains in projection (this study).
4.3.4 Extraction of grain boundary parameters
The grain boundaries of crystals with curved, complex or sutured surfaces can
be parametrised in a number of ways. Although grain boundaries are threedimensional structures, so far they have only been investigated in detail in 2-D
sections. The stereology of conversion of such 2-D measurements to 3-D
parameters has not generally been investigated. The simplest descriptor of
the shape of the grain is the ‘Haywood circular parameter’ which is the ratio of
the actual perimeter to that of a circle of equal area. This parameter is more
sensitive to the overall axial ratio of the grains (see Section 4.3.3) rather than
the irregularities in the outline, which we seek here. The shape of loose rock
particles is very important in sedimentology and hence the methodology is well
established, but outside the scope of this volume (e.g. Barrett, 1980, Lewis &
McConchie, 1994a, Lewis & McConchie, 1994b).
150
Grain shape
4.3.4.1 Fractal analysis of grain shapes
The shape of a grain boundary can be characterised by fractal methods
(Mandelbrot, 1982, Orford & Whalley, 1983, Kruhl & Nega, 1996, Carey
et al., 2000). So far such methods have only been applied to sections or
projections of clasts with the assumption that such measurements also characterise the parent 3-D object. A smooth surface (in a section) will have a
fractal dimension of one, which will increase to a maximum of two for a planefilling line. Outlines of particles of geological interest tend to have more
restrained fractal dimensions from 1 to 1.36. However, it should be noted
that this fractal dimension is that of a section through a surface, which will
have a higher fractal dimension of between 2 and 3.
The fractal dimension of an outline is generally determined using the divider
or box counting methods (Figure 4.10) and the results presented on
Mandelbrot–Richardson (log–log) diagram. A single straight line on an M–R
diagram indicates monofractal behaviour. That is, the shape is self-similar at
all of the scales measured. Many objects of geological interest give multiple
straight-line segments on an M–R diagram, which is termed multifractal
behaviour. Some authors have distinguished two scales and others have
divided up the fractal into several size segments (e.g. 3–30 pixels; 31–60 pixels;
etc.), that were then analysed using multivariate statistics (Dellino & Liotino,
2002). Curved distributions on M–R diagrams indicate nonfractal behaviour
and should not be forced into line segments. Care must be taken when outlines
(a) Divider method
(b) Box counting method
13 divider steps
29 boxes intersected
Figure 4.10 Methods for determining the fractal dimension of the outline of
a grain. (a) Divider method. A line of length l is walked around the outline
and the number of steps counted. This is repeated for different lengths and a
graph of log(n) against log(l) is plotted (Mandelbrot–Richardson diagram).
The slope is the fractal dimension of the line. (b) Box counting method. A box
size of length l is defined and the fraction of boxes, n, that contain a part of the
outline are counted. This is repeated for different box sizes and a graph
plotted of log(n) against log(l). The slope is the fractal dimension.
4.3 Methodology
151
are measured: the resolution, that is number of pixels, provides a lower size
limit for the method. Also in many outlines the undulations are sufficiently
infrequent that the starting position for the divider method affects the results.
Several different positions and directions can be averaged to improve the
resolution (Dellino & Liotino, 2002).
4.3.4.2 Gaussian diffusion–resample method
Saiki et al. (1997, 2003) have proposed a new rounding parameter that is not
dissimilar to fractal methods. A selection of grains of similar size is first
reduced to a single binary image (black and white) and the total area noted.
The image is then passed through a Gaussian diffusion filter with an aperture
of 0.5 pixels. The outlines are now blurred. The image is then rebinarised with a
threshold such that the area of the grains is unchanged. The original and final
images are subtracted so that the material eroded by the process can be
measured. The process is repeated with a wider Gaussian filter. The eroded
areas are normalised to the total area and plotted against the aperture: the
slope of the line is a measure of the irregularity of the outline.
4.3.4.3 Fourier analysis of grain shapes
The overall shape of a grain intersection or outline can be evaluated precisely
at all scales by the use of Fourier analysis (Ehrlich & Weinberg, 1970).
However, data analysis methods may restrict the scale of application: most
individual applications of a method cannot resolve the outline at a scale better
than 0.1 to 0.01 times the size of the outline. More detailed measurements
may be ‘stitched together’ from several analyses to overcome this problem.
It should be also remembered that Fourier analysis is generally applied to only
a single section through a three-dimensional object. Three-dimensional
Fourier analysis is possible, but has not yet been done on geological materials.
The radius of the grain outline at angle Y, RðYÞ is traced out. These data
are transformed by classical Fourier analysis into a series of harmonics
(Figure 4.11). The basic shape is a circle of equivalent area; the first harmonic
is an ellipse and gives the elongation. The second harmonic gives a measure of
triangularity and the third gives squareness. The harmonics continue up to the
number of data points. The shape can be reconstructed as:
RðYÞ ¼ a0 þ
N
X
ðan cos n Y þ bn sin n YÞ
n¼1
where N ¼ the total number of harmonics, n ¼ harmonic number, and a and b
are coefficients that give the magnitude and phase of each harmonic. The
152
Grain shape
Y
Θ)
Θ
R(
0
1
X
3
2
Figure 4.11 In classical Fourier analysis the radius of the outline, R, is
decomposed into a basic circle (n ¼ 0) and a series of harmonics, 1, 2, 3, . . ., n.
Y
x8,y8
x7,y7
x2,y2
x6,y6
x5,y5
x8,y8
x2,y2
x7,y7
Re-entrant
x3,y3
R2(Θ)
x4,y4
R1(Θ)
Θ
x6,y6
X
x3,y3
x5,y5
Figure 4.12
x1,y1
x1,y1
x4,y4
Re-entrant problem in classical Fourier analysis.
method is useful because it gives a measure of the shape of the profile at all
scales, covering the field of sphericity, roundness and grain boundary shape
(surface texture). There is, however, a problem for grains with re-entrants:
there can be two radiuses for certain angles (Figure 4.12). This is not usually a
problem for mature sediments, but occurs in volcanic fragments and grains in
plutonic or volcanic rocks.
An approach that avoids this problem is the Fourier Descriptors method
(Thomas et al., 1995, Bowman et al., 2001). In essence Fourier analysis is
applied to the x, y coordinates of the outline of the grain and not its radius
(Figure 4.12). The method uses complex numbers, with the x coordinate as the
real part of a number and the y coordinate as the imaginary part. The outline is
4.3 Methodology
153
sampled at regular intervals to give a series of points xm, ym. The outline can be
reconstructed from:
þN=2
X
2nm
2nm
xm þ iym ¼
ðan þ i bn Þ cos
þ i sin
M
M
n¼N=2þ1
where N is the total number of descriptors, n is the descriptor number, M is the
total number of points, m is the index number of each point and a and b are
coefficients for each descriptor (i denotes the imaginary part of a complex
number). The descriptors n ¼ 0, 1, 2, 3 give the radius (equivalent to
size), elongation, triangularity and squareness as in classical Fourier analysis.
The descriptors þ1, þ2, þ3 give asymmetry, second-order triangularity,
second-order squareness (Bowman et al., 2001). The descriptors give almost
too much information on shape. We are generally concerned more with the
distribution of roughness at different scales. Drolon et al. (2003) have proposed that the Fourier descriptors are recast using wavelet techniques into a
series of multiscale roughness descriptors. These give a measure of the total
‘energy’ at each scale. They appear to offer a simpler way of looking at
roughness that is better at classifying sedimentary grains.
4.3.4.4 Shape parameters used in sedimentology
Some studies of crystalline rocks have used sedimentological shape parameters. One such measure is the circularity. This is defined by the perimeter
of the grain outline:
circularity ¼ 4ðA=P2 Þ
where A is the area and P is the perimeter of the grain outline. It varies from
one for a perfect circle to zero for an infinitely elongated grain outline. The
elongation of a grain has several definitions: in 3-D it is the ratio of the longest
and shortest axes. In 2-D it is the aspect ratio of its outline: that is, intersection
length/width.
The grain shape can also be evaluated at a smaller scale by quantifying the
roundness. This is usually measured from grain intersections using the
Waddell formula:
roundness ¼ ðr=RÞ=N
where r is the radius at the grain corners, R is the largest inscribed circle (circle
enclosed in the grain outline) and N is the number of corners (Figure 4.13).
This formula is rather difficult to apply to real grains; hence roundness is
estimated from silhouette diagrams.
154
Grain shape
r2
r1
r3
R
r4
r5
Figure 4.13 The ‘roundness’ parameter of grains, as determined from a
section through the centre of a grain.
4.3.5 Determination of dihedral angles
Dihedral angles are a 3-D property and can be measured directly in transparent materials using a universal stage (see Section 5.5.1). The sample is rotated
so that all three grain boundaries are vertical and the dihedral angles measured
directly (Vernon, 1970) or using a stereographic plot (Vernon, 1970). It is also
possible to use the tilting capabilities of a transmission electron microscope to
examine true dihedral angles in thin films (Cmiral et al., 1998). However, in
most studies the apparent dihedral angles are measured in 2-D sections, such
as thin sections and plotted on cumulative frequency diagrams. If there is only
one true dihedral angle in a population then it is close to the median of
apparent dihedral angles (Riegger & van Vlack, 1960). However, if there are
a range of true dihedral angles, then this range cannot be determined uniquely
from the population of apparent (2-D) dihedral angles (Jurewicz & Jurewicz,
1986, Elliott et al., 1997). The overall distribution of apparent dihedral angles
has been used to determine if a monomineralic rock is equilibrated or not
(Elliott et al., 1997). Impingement textures can clearly be distinguished from
equilibrium textures (Figure 4.5), but the method is not sensitive to small
variations in the degree of equilibration. Hence, it is better to use 3-D measurements where possible.
4.4 Typical applications
4.4.1 Undeformed crystalline rocks
The development of crystal faces in minerals growing unconfined in silicate,
aqueous or other fluids has been of interest to mineralogists since the
155
4.4 Typical applications
nineteenth century. The monumental work of Dana et al. (1944) compiled
the faces developed in a wide variety of minerals, with emphasis on the nonsilicates. A more recent and comprehensive survey is that of Kostov and
Kostov (1999). They give much more detail and applications of habit to
mineral exploration and other aspects of geology. However, apart from
plagioclase and diamond there are few quantitative studies of crystal shape.
4.4.1.1 Plagioclase
Plagioclase is the most abundant mineral in the crust; hence there is much
interest in any petrographic parameter that can reveal more of its petrogenesis.
In addition, plagioclase crystal size distribution has been studied in both
volcanic and plutonic rocks and crystal shape is commonly determined in
such projects (Figure 4.14). At first glance, controls on crystals shapes seem
to be rather different in the microlites of volcanic rocks, as opposed to volcanic
phenocrysts and plutonic crystals.
Crystallisation of microlites in volcanic rocks can increase gas pressure and
initiate eruptions (Sparks, 1997, Cashman & Blundy, 2000) Hence, there has
been much quantitative work on microlites. Experimental evidence suggests
that there is a progression from tabular crystals at low undercooling (30 8C)
through acicular habits and finally to skeletal crystals (Lofgren, 1974). The
1.0
Volcanic rocks
Plutonic rocks
short / intermediate axes
0.8
0.6
0.4
0.2
0
0.0001
0.001
0.01
0.1
1
log (characteristic length) (mm)
10
Figure 4.14 Short/intermediate axial ratios of plagioclase against
characteristic or typical crystal length in volcanic and plutonic rocks
(compilation of data cited here and in Section 3.4). Points from the same
unit are linked. All crystals are close to tabular in form except for the open
diamond, which is prismatic. The open squares are for foliated plutonic
rocks.
156
Grain shape
shape of natural plagioclase microlites has been recorded frequently and many
authors have noted that the shape of plagioclase microlites varies with their
size: Cashman (1992) noted changes from an aspect ratio of 1:8 in the 1980 Mt
St Helens blast dacite (typical length 10 mm) to 1:5 and finally 1:3 in the later
erupted rocks (typical length 100 mm). Examination of her thin section photographs shows that the crystals are tabular. Higgins (1996b) found that crystals
smaller than 200 mm in outline length have an aspect ratio of 1:5:6 whereas
phenocrysts had a ratio of 1:3:4 (Figure 4.14). In both these cases the crystals
are tabular: twin-plane orientations, where visible, indicate that the tabular
face is parallel to {010}. Hammer et al. (1999) observed that the smallest
microlites in a dacite are tabular (0.7 mm; 1:3:3), whereas the larger microlites
are prismatic (1.6 mm; 1:1:7). They ascribed this to a change from a nucleation
dominated crystallisation regime to one dominated by growth. These microlites are much smaller than those measured in other studies, hence there may be
more complexity at this scale which is not seen in the larger microlites.
Compilation of plagioclase shapes from numerous sources reveals a broad
correlation between the characteristic size of plagioclase crystals in volcanic
rocks and their shape, as expressed by the short/intermediate ratios, with the
exception of the smallest microlites (Figure 4.14). However, not all rocks have
the same trend; hence it is possible that both size and shape track another
parameter and the most likely is undercooling. Initially, a high degree of
undercooling is required to initiate nucleation. Under such conditions the
growth rate of the tablet edge is much greater than that of the tablet side. As
crystal growth occurs, release of latent heat reduces the undercooling and the
ratio of growth rates is reduced, hence leading to more equant crystals.
Plagioclase in plutonic rocks, such as anorthosite, gabbro, diorite and
granodiorite does not generally have euhedral faces; however it commonly
has a distinctive habit that depends on rock type and conditions of crystallisation (e.g. Higgins, 1991). Higgins (1991) examined plagioclase crystals from
both massive and well-foliated anorthosites from the Sept Iles mafic intrusion
and noted a significant difference in plagioclase shape. In the massive anorthosite the plagioclase had an overall axial ratio of 1:2:4, whereas in the laminated
anorthosite the plagioclase was considerably more tabular with an axial ratio
of 1:5:8 (the shapes shown here were recalculated from the original data). This
suggests that growth under conditions of magma shear promotes the growth
rate of the tablet edges (Higgins, 1991). In another study of plagioclase in
anorthosite the crystal shape changed with coarsening from tabular crystals to
more equant forms (Higgins, 1998).
It is possible that the crystal shape in both volcanic and plutonic rocks just
reflects the chemical potential gradient around the growing crystal: where
157
4.4 Typical applications
)
(001
)
growth rate
(010)
(010)
(001
Rate(001)
Rate(010)
chemical potential gradient
Figure 4.15 Possible response of plagioclase face growth rates to changes in
the chemical potential gradient around the growing crystal (this study). The
(001) face is taken to be typical of the tablet edge faces. The chemical potential
gradient will be high at high degrees of undercooling as diffusion will be
unable to supply nutrients to the crystal fast enough. The chemical potential
gradient will be lower at lower undercooling where the crystals are growing
more slowly or where shear advection removes depleted magma from around
the crystal faces.
there is a high gradient the tablet edges are favoured, but under more equilibrium conditions the effect is reduced and growth is more equal on the tablet
sides and faces (Figure 4.15).
4.4.1.2 Zircon
Zircon is a widespread, chemically resistant mineral that is used for geochronology and stable isotope studies (Hanchar & Hoskin, 2003). Zircon is not
abundant in most rocks; hence crystals are generally extracted mechanically
before analysis. The problem of extracted crystals is that they lose their
petrographic context. However, zircons are tough and many retain their
habit after extraction. Qualitative and quantitative studies of zircon habit
has therefore received much attention (Figure 4.16; Pupin, 1980, Kostov &
Kostov, 1999).
Zircons in rhyolite lavas and tuffs are prismatic and their length/breadth
ratio varies from 1.3–1.5 for crystals 50 mm long to 2.1–3.0 for crystals 250 mm
long (Bindeman & Valley, 2001, Bindeman, 2003). This may be an intrinsic
property of zircons (Larsen & Poldervaart, 1957). Vavra (1993) has examined
quantitatively the growth of individual zircons using both their overall shape
158
Grain shape
Zircon crystal forms
High temperature
Dry magma
101
101
100
100
Bipyramidal
101
100
Acicular
101
110
110
Equant
Low supersaturation
High alkalinity
High Th, U, REE
Figure 4.16
Prismatic
Low temperature
Hydrous magma
High supersaturation
Low alkalinity
Low Th, U, REE
Controls on zircon morphology (Kostov & Kostov, 1999).
and internal zoning patterns. He again finds that the pyramidal faces in general
grow faster than the prismatic faces as the undercooling (crystallisation driving force) increases, hence producing the same change in shape with crystal size
that was observed qualitatively by other authors.
4.4.1.3 Other minerals
Euhedral quartz crystals commonly grow in hydrothermal metamorphic
environments as prismatic crystals with pyramidal terminations. Ihinger and
Zink (2000) used infrared spectroscopy to show how such crystals develop.
They found that as the crystal grew the termination evolved from one dominated by {11
11} faces to a final form with {1011} faces. They showed that an
abundance of impurity phases such as HOH and AlOH can be used to measure
the relative growth rates of the faces. However, it is not clear if the change in
the relative growth rates of the faces was related to an increase in size or
changing crystallisation conditions, such as temperature.
Crystals with rapid growth textures are found in many rocks and the most
commonly described is olivine (Faure et al., 2003). Up to ten different types of
olivine morphology have been described (Donaldson, 1976) and these are
broadly related to undercooling (supersaturation). The experiments of Faure
et al. (2003) have shown that there are only three true morphologies and other
observed shapes are just special sections. With increasing undercooling olivine
159
frequency
frequency
4.4 Typical applications
0
10
20
length (μm)
30
0
1
2
3
width (μm)
4
5
Figure 4.17 Shape of pyroxene microlites in rhyolitic obsidian (Castro et al.,
2003). The length of the crystals is much more variable then their width.
morphology went from tablets to hopper (skeletal) to swallowtail (dendritic).
The development of the different forms is a function of both cooling rate and
degree of undercooling.
Castro et al. (2003) have examined pyroxene microlites in rhyolitic obsidian
using an innovative optical sectioning technique (see Section 2.2.2). They find
that the crystal lengths vary over a factor of ten (3–30 mm), but the widths only
vary by a factor of two (Figure 4.17). These pyroxene crystals initially develop
as stubby prisms, with a width of 1–2 mm, and then just grow along the
prismatic axis. Hence, the aspect ratio is proportional to the length. This
behaviour can be produced if the ratio of the growth rates of prismatic faces
over pyramidal faces decreases with undercooling.
Zeh (2004) investigated the shape of apatite grains in a series of metamorphic rocks. He found that the ratio of the length/diameter (l/d ) of the
crystals and mean crystal length varied systematically with temperature of
metamorphism. Those crystals from the lowest grade rocks had a mean l/d of
20 and a length of 1 mm and those from the highest grade had a mean l/d of 2
and a length of 100 mm. Allanite showed a similar change in shape, but it was
much less pronounced.
Pallasites are unusual meteorites that are mostly comprised of olivine crystals set in a matrix of Fe–Ni alloys. In some pallasites the olivine crystals are
euhedral or angular fragments, whereas in others they are well rounded. Saiki
et al. (2003) showed experimentally that rounding of olivine could occur in
geologically reasonable periods of time at a temperature of 1400 8C, which is
sub-solidus for these compositions. Adjustments to the shape of the olivine
grains occurred by volume diffusion in the crystal lattice and not by surface
diffusion along the grain boundaries.
160
Grain shape
4.4.2 Deformed rocks
4.4.2.1 Sutured quartz grain boundaries
Sutured grain boundaries have been observed in many deformed rocks, but
seem to be particularly evident on quartz–quartz boundaries (Figure 4.18).
Where examined in thin section such grain boundaries are scale invariant over
a length scale of one to two orders of magnitude (Kruhl & Nega, 1996). The
fractal dimension varies from 1.05 to 1.30, but it should be remembered that
this is the value for an intersection of the grain boundary with the plane of the
thin section. The 3-D fractal dimension is not known, but should be related to
these values.
During deformation at high temperatures there is equilibrium between the
production of additional undulations of the grain surface and their removal by
coarsening (ripening) processes as the texture approaches equilibrium. Hence
the degree of suturing can be related to the strain rate of the rocks and their
temperature. Kruhl and Nega (1996) analysed three disparate groups of rocks
and proposed that the fractal dimension (D) can be used as a geothermometer:
At a temperature of 350 8C D ¼ 1.28 and at 700 8C D ¼ 1.08. However, the
strain rate is also a controlling factor (Takahashi & Nagahama, 2000). In
addition, overprinting by subsequent events can produce scaling heterogeneities (Suteanu & Kruhl, 2002). Clearly, the application of this technique as a
geothermometer, although useful, requires careful observation of other
parameters.
Detailed examination of sutured quartz–quartz boundaries using a universal
stage has shown that they are curved, but are composed of multiple straight
segments (Kruhl & Peternell, 2002). The orientation of these crystal faces is
350 °C D = 1.28
700 °C D = 1.08
0.5 mm
Figure 4.18 Tracing of two sutured quartz–quartz grain boundaries (Kruhl
& Nega, 1996). The outer curve was developed at 350 8C and has a fractal
dimension of 1.28, whereas the inner curve developed at 700 8C and has a
fractal dimension of 1.08.
4.4 Typical applications
161
controlled by the lattice orientation of the crystal and its neighbour. At low
temperatures (350 8C) the faces conform to a wide variety of indices, but as
the temperature increases to 700 8C faces close to {1011} are favoured. The
presence of the crystallographically controlled faces probably makes the texture more resistant to addition changes, until the rock is deformed again.
4.4.2.2 Calcite in mylonites
Amongst the vast number of articles on the orientation of crystals in deformed
rocks, there are only a few in which the shape of the crystals have been
determined. Clearly, this can give information on deformation, like the size
and orientation of the crystals. Herwegh et al. (2005) examined a series of
carbonate mylonites from the Alps that formed at temperatures from 250 to
380 8C. They found that crystals from the low-temperature mylonites had a
ratio of S/I of 0.8, which decreased to 0.45 for the highest temperatures. They
also used the Paris shape factor to determine the sinuosity of the crystal
outlines and found that it, too, increased with temperature. This seems to be
the inverse of what was found in quartz mylonites (see above).
4.4.3 Dihedral angles
Dihedral angles can be used to explore the flow transport properties of fluidbearing materials, hence there has been considerable experimental work in this
field, but this has not been balanced by study of natural materials. In addition,
there is a problem concerning the measurement methods: many early studies
used apparent dihedral angles measured in sections that are difficult to assess
in the light of the stereological studies of Elliott et al. (1997). Although 2-D
studies form the bulk of the published work, such methods should only be used
where 3-D methods cannot be applied, such as opaque minerals and small
samples. Finally, there is some ambiguity in the way that dihedral angles are
reported for polyphase rocks. Some authors record DAs as plagioclase–
plagioclase–clinopyroxene: it is clear here that the DA of the last phase,
clinopyroxene, is measured. However, others use terms like melt-solid-solid,
in which the DA of the first term, the melt, is measured. There is little
ambiguity for two phases, but for three phases the phase in which the DA is
measured must be specified. I use here the convention A(B-C) to indicate that
the DA was measured in A at a contact with B and C.
The earliest studies of dihedral angles in natural rocks were in granulites
(Kretz, 1966b, Vernon, 1968, Vernon, 1970). Clinopyroxene (olivine–olivine)
and clinopyroxene (plagioclase–plagioclase) intersections both have median
dihedral angles of 1158 178 (1 sigma). This indicates that there is little
162
Grain shape
contrast between the interfacial energies of these three minerals under these
conditions.
Many earlier DA studies were centred around three themes: basaltic melts
in peridotites, partial melting of crustal silicate rocks and H2O–CO2 crustal
fluids. Commonly, these studies were experimental with relatively few studies
of actual rocks. The distribution of melt in peridotite is of interest for the origin
of basalt by partial melting (e.g. Faul, 1997) in addition to the geophysical
properties of the mantle, such as seismic attenuation and mechanical strength.
The DAs of basalt (olivine) and other minerals have been investigated in both
experimental and natural materials (e.g. Waff & Bulau, 1979, Cmiral et al.,
1998, Duyster & Stockhert, 2001). In general, olivine crystals surrounding melt
pockets are faceted, not curved, but the dihedral angles are low to zero (Cmiral
et al., 1998). Other minerals in mantle rocks have a significant effect on the
overall permeability: diopside and spinel increase permeability whereas enstatite lowers it (Schafer & Foley, 2002). Hence, melts should flow more easily
through spinel lherzolites than harzburgites.
Many types of granite originate by partial fusion of the crust, followed by
aggregation and escape of the magma. Laporte and Watson (1995) have
examined the process using experimental and theoretical techniques, with
special emphasis on the role of biotite and amphibole. They concluded that
crustal rocks cannot be modelled as ideal systems because they are comprised
of many different minerals, mostly with medium to high degrees of anisotropy
of interfacial energy. The overall fabric (crystallographic preferred orientations) and heterogeneity of the rock also had an important effect. For rocks
with highly anisotropic minerals, such as biotite, and low degrees of melting,
the silicate liquid lies in isolated plane-faced pocket at grain junctions. The
shape of these pockets is controlled by the anisotropy of the minerals and the
texture of the rock. The low DA for liquid (solid–solid) junctions in crustal
rocks indicates that melts should be connected at very small degrees of melting
and hence should be continuously drained from a rock during melting.
Amphibolites may have a connectivity threshold of 3%. However, such
melts are very viscous and hence viscosity and not connectivity may limit the
separation of melt and restite. In contrast, well-foliated biotite rich rocks may
act as barriers to transverse liquid flow.
The role of fluids, particularly those rich in water, in geological processes is
pivotal. Holness (1993) examined experimentally the behaviour of H2O–CO2
fluids in quartzites. Water-rich fluids have DAs close to or lower than the
critical value of 608 at both low and high temperatures, which was attributed to
the presence of a film of water on the quartz surface. This implies that there is
a window of permeability at both low and high temperatures for infiltration
4.4 Typical applications
163
of water. The behaviour of CO2 was completely different: it had high DAs at
all temperatures suggesting that it was not absorbed onto the quartz. The presence
of solutes like NaCl appears to reduce further the DA of aqueous solutions and
hence enlarge the window of permeability (Watson & Brenan, 1987).
Although processes during the early stages of accumulation of crystals in
magma chambers are reasonably well understood, there is little information
on the later textural evolution of these rocks. On the basis of crystal size
distributions Higgins (2002b) and Boorman et al. (2004) both considered
that textural evidence for crystal accumulation is frequently overprinted by
coarsening. Holness et al. (2005) have attacked the problem by the analysis of
DAs. They pointed out that initial growth should generate impingement
textures with a median dihedral angle of about 608 (Figure 4.5). Analysis of
melt (clinopyroxene–clinopyroxene), melt (amphibole–amphibole) or melt
(plagioclase–plagioclase) DAs in glassy lavas established a range of median
DAs from the impingement values down to 288. For the Kula lavas DAs and
changes in crystal shape indicated that textural equilibration was achieved.
However, for the Kameni lava they favoured diffusion limited growth.
Cumulate plutonic rocks do not contain any melt, but some minerals may
pseudomorph former melt pockets. Hence, the next step was to examine the
DAs of such minerals in cumulate rocks.
The small layered intrusion on Rum, Scotland has received enormous
attention, but there are still major uncertainties in its petrogenesis: one such
problem is the origin of peridotite layers. These may have formed as accumulations of olivine or by injections of olivine-rich picrite into the semi-solid
layered gabbros and troctolites. Holness (2005) has investigated the textures of
the rocks that host the peridotite layers to determine their thermal history and
hence resolve their origin. She considered that clinopyroxene pseudomorphs
the last melt and measured the clinopyroxene (plagioclase–plagioclase) (CPP)
DAs along a section orthogonal to a peridotite layer. She found that all CPP
median DAs lay between the values for impingement textures (608) and the
values for complete equilibrium (115–1208, Figure 4.19). It is assumed that
higher median values reflect movement towards solid-state equilibration during reheating by the injection of peridotite sills. In most of the traverses there
was no correlation between median DAs and proximity to peridotite layers,
except for one layer. Hence, both models for the origin of the peridotite layers
are validated, but for different layers.
Sulphide minerals commonly occur as globules in silicate rocks. Mungall
and Su (2005) showed experimentally that this was because of the high surface
energy between sulphide and silicate liquids, as compared to that between
silicate minerals and silicate liquids. They considered that at low degrees of
164
Grain shape
35
Impingement
standard deviation
30
Kula, Santorini melt(cpx-cpx);
melt(amph-amph) and melt(plag-plag)
Rum, cpx(plag-plag)
Equilibrium, cpx(plag-plag)
25
20
15
10
5
20
Solid-state
equilibrium
Melt-present
equilibrium
40
60
80
median angle (°)
100
120
Figure 4.19 Median against standard deviation of dihedral angles in meltpresent and solid materials (Holness et al., 2005). The Kula and Kameni data
(squares) are from partially melted crystalline enclaves. The data from Rum,
Scotland (filled circles) are from solid gabbro and troctolite (Holness, 2005).
Granulite samples (empty circles) provided data for solid-state equilibrium
(Kretz, 1966b, Vernon, 1968, Vernon, 1970).
supersaturation sulphide droplets may nucleate at wide intervals leading to
kinetic control of composition. The high surface energy of sulphide droplets
will hinder their migration through a crystal mush, unless propelled by moving
silicate liquids. Clearly, this has important implications for the origin of
massive sulphide bodies.
Notes
1. The terms crystal habit, crystal shape and crystal morphology are considered to be
synonyms.
5
Grain orientations: rock fabric
5.1 Introduction
The quantitative study of grain orientations (fabric)1 is very important in
structural geology and materials science (Wenk, 1985, Kocks et al., 2000,
Randle & Engler, 2000, Wenk, 2002, Wenk & Van Houtte, 2004). It is less
established in igneous petrology (Nicolas, 1992, Smith, 2002), where fabric
has been used mainly to establish flow directions and mechanisms in lava
flows and dykes. Fabric studies find their way in geophysics as there is
considerable interest in the anisotropy of seismic velocities (e.g. Mainprice
& Nicolas, 1989, Xie et al., 2003). Fabric is also important in engineering
geology, as it affects the physical properties of rocks (e.g. Akesson et al.,
2003). It is also of interest in palaeontology and medicine for quantifying
the structure of bone (Ketcham & Ryan, 2004). Knowledge of grain
orientations is also necessary to calculate size distributions from the dimensions of grain intersections in all the domains already mentioned (see
Section 3.3.4).
The grains that define the fabric of a material can be a variety of different
objects: commonly crystals, clasts and enclaves. One measure of fabric is based
on the orientations of the grain shapes (shape preferred orientations: SPO).
Here the nature of the material is unimportant for the measurement of the
fabric. However, the contrast in viscosity between the grain and matrix during
production of the fabric will define what parameter is recorded by the fabric.
For example, in magma originally spherical deformable enclaves will record
strain, whereas rigid crystals will rotate and just record indirectly aspects of the
deformation.
If the grains studied are crystals then another measure is possible: the fabric
can be defined by the orientation of the crystal lattices, here termed lattice
preferred orientations or LPO (also known as crystallographic preferred
165
166
Grain orientations: rock fabric
orientations; macrotexture or texture in materials science literature). Of course,
if crystals are euhedral then these two orientations are parallel. This is generally
not the case in most metamorphic and many plutonic rocks, as crystals do not
grow in an unobstructed environment or they may have been partly dissolved.
In this situation orientations defined by crystal shape and lattice will not
necessarily be related.
5.2 Brief review of theory
5.2.1 Fabrics developed during growth of new crystals
5.2.1.1 Crystallisation from a magma
Isolated crystals in a homogeneous liquid nucleate with random orientations.
If there is no flow of the magma, then growth will conserve the orientation of
the nuclei and there will be no development of fabric. However, if the growth
of crystals is constrained by the presence of other crystals, or if there is a
significant concentration gradient in the magma then a fabric may develop
during crystallisation. Such fabrics are growth phenomena, hence they are
both shape and lattice preferred orientation fabrics.
The most common type of crystallisation fabric in plutonic rocks is comblayering (Lofgren & Donaldson, 1975). This texture is produced where
crystals initially nucleate on a rigid or semi-rigid plane. The orientation of
the nuclei is not important. If the crystals have an initial shape anisotropy
then competition between the crystals will eliminate those with maximum
growth rates parallel to the crystallisation plane. This leads to a strong linear
fabric, normal to any structures that define the original solidification front,
such as mineral layering. Crystals may become curved or branching if they
can grow fast enough to escape the zone depleted in nutrients close to the
solidification front. Similar fabrics can be developed in some volcanic rocks:
the most spectacular are the spinifex textures in komaiites (e.g. Shore &
Fowler, 1999).
Equilibration of magmas or rocks is the process by which the surface and
internal energy of the system is minimised. One process is by the coalescence of
mineral grains, which reduces the total surface energy. Energy is further
minimised if the differences between the orientations of the crystals is reduced
by rotation of grains. There is evidence for this in the commonly observed
phenomena of synneusis of igneous minerals such as olivine and plagioclase
(see Section 6.2.3; e.g. Schwindinger & Anderson, 1989). Recently evidence has
been found that some garnet porphyroblasts are also formed by coalescence
and rotation of many grains (Prior et al., 2002).
5.2 Brief review of theory
167
5.2.1.2 Solid-state mineral crystallisation
When crystals nucleate and grow in a solid medium then the orientation of the
crystal will be organised so as to minimise the energy required for growth.
Hence, growth of crystals (or transformation of existing minerals) will generate fabrics that are defined by both the crystal lattice and shape. If deformation is coaxial then SPO and LPO will be parallel. Under non-coaxial
deformation the obliquity between the LPO and SPO can be used to determine
the sense of shear. Unless the deformation that lent the crystals their orientation ceases immediately afterwards the newly formed crystals will be deformed
in the solid state as they grow.
5.2.2 Fabrics produced during magmatic and solid-state deformation
of existing crystals
5.2.2.1 Property changes during solidification
The solidification of magma and the melting of rock, are not simple processes,
but involve a transition in properties from a liquid to a solid. Three stages are
usually recognised but the transition points appear to be different for solidification and melting (e.g. Vigneresse et al., 1996, Arbaret et al., 2000). During
solidification from 0 to 30%–40% crystals the magma behaves as a suspension,
with pure Newtonian behaviour, especially at low crystal concentrations. Above
30%–40% crystals interact frequently and the magma acquires yield strength.
The magma can now be considered to behave as a Bingham fluid. A rigid
framework of crystal chains, or more accurately foam-like crystal surfaces,
may develop in some rocks at this stage, or even earlier (Philpotts et al., 1999).
Otherwise, the quantity of crystals must increase to 50%–65% crystals before a
continuous network of crystals forms and the yield strength increases drastically. The magma then has a plastic behaviour that is close to that of a solid. At
this point strain may be concentrated into discrete zones, crystals may fracture
or be divided into sub-grains and folding may occur. Clearly the first stage must
be considered to be in the igneous domain. The last stage is more metamorphic,
but the intermediate stage is difficult to classify. In many plutonic rocks deformation continues through all these stages maintaining the same orientation.
During melting the rock becomes much weaker with as little as 5% melt
(95% crystals) as melt pockets become interconnected (Rosenberg & Handy,
2005). A second major rheological change occurs about 30% melt after which
the material resembles magma. The differences between solidification and
melting probably reflect textural changes: for instance during melting small
crystals will be removed first, leaving isolated larger crystals. Whereas during
168
Grain orientations: rock fabric
Simple shear flow
Hyperbolic shear flow
Pure shear flow
Figure 5.1 Flow lines for different shear flow types. Simple and hyperbolic
shear are non-coaxial shear; pure shear, or compaction, is coaxial shear.
solidification small crystal will nucleate and grow as long as significant undercooling is maintained.
Deformation can occur in both rocks and magmas. Plastic and solid-state
deformation of crystals not only changes the shape of crystals, but also their
size. The basic theory of these processes is discussed in Section 3.2.5. Magmatic
deformation will only change the orientation and position of crystals, but it
depends on the type of shear regime. Shear can be classified with reference to a
fixed plane (Figure 5.1). Simple shear is parallel to the plane, pure shear is
normal to it and hyperbolic, or mixed, is between the two extremes.
5.2.2.2 Magmatic deformation
Numerical models of magmatic deformation are based on the motions of rigid
particles in a fluid. There are a number of assumptions (Arbaret et al., 2000,
Iezzi & Ventura, 2002). (1) Particles are ellipsoids or parallelepipeds of varying
dimensions. (2) They do not interact with each other. (3) They do not change
shape during deformation. (4) The liquid is Newtonian and flows constantly
without turbulence. The most recent numerical modelling was done in three
dimensions using particles of different shapes and shear regimes (Iezzi &
Ventura, 2002). Under simple and hyperbolic shear, cubes and elongated
tablets never achieved a stable configuration: they just continued to roll
(Figure 5.2a). Simple 2-D analogue modelling shows that at high concentrations of particles interactions between adjacent grains prevents cyclical movements and grains become imbricated (Ildefonse et al., 1992). Tablets and
prisms rotated towards the plane of shear and produced a strong fabric
(Iezzi & Ventura, 2002). For pure shear (compaction) all shapes rotated until
they were parallel to the shear plane (Figure 5.2b). However, this process is not
very efficient and high degrees of deformation are necessary to produce a
visible fabric (Higgins, 1991, Ildefonse et al., 1992).
The lack of stability of fabrics for grains with equant or near-equant shapes
means that local fabrics may not indicate the direction of shear. Iezzi and
169
5.2 Brief review of theory
(a) Simple shear
Figure 5.2
(b) Pure shear
Schematic behaviour of a cube and tablet in a sheared magma.
Dextral
Figure 5.3 Imbrication or tiling of grains or crystals for dextral simple
shear. Rotation of individual crystals in response to simple shear is stopped
by the presence of an adjoining crystal.
Ventura (2002) showed that the strain regime can only be determined precisely
if the orientations of crystals with different shapes can be determined.
Normally this would be for different minerals.
Mineral grains do usually interact, especially at higher grain concentrations,
but this has not been modelled numerically yet. One result of such interactions
is tiling or imbrication (Figure 5.3; e.g. Blumenfeld & Bouchez, 1988). This is
when the rotation of one grain is stopped by impact against another grain.
Two-dimensional analogue models indicate that tiling can be used to identify
shear directions and planes (Ildefonse et al., 1992). Nicolas and Ildefonse,
(1996) have proposed that crystal mushes in magma chambers flow and
develop fabrics after deposition without plastic deformation. They suggest
that magmatic flow of crystal-rich material (>80% crystals) is enabled by
pressure-solution at imbrication points: flow will be stopped until sufficient
material has been removed and the crystals are no longer imbricated. Flow
then resumes until imbrication is again encountered.
Magmatic deformation is controlled by the shape of the grains; hence the
fabric is the result of SPO. If the crystals are the results of growth without
significant solution, then their shape will reflect the orientation of the crystal
lattice and the LPO will be parallel to the SPO.
170
Grain orientations: rock fabric
Many magmatic rocks contain enclaves – generally rounded patches of
texturally or compositionally contrasting rocks. Enclaves are generally
assumed to result from the mixing of two or more magmas. If the enclaves
have a significantly higher liquidus temperature than the magma then the
enclaves will be chilled and act as rigid bodies during magmatic deformation,
similar to the crystals discussed above. However, if the compositional contrast
is small then there will be little viscosity contrast between the enclaves and the
magma. If the enclaves were originally spherical then their shape will record
the total magmatic strain.
5.2.3 Influence of coarsening and pressure solution compaction on fabric
Coarsening (Ostwald ripening, annealing, textural maturation) is the process
by which the energy associated with the surface of the crystals in a rock is
minimised (see Section 3.2.4): Small crystals dissolve at the same time as large
crystals grow. This process may be important in many igneous rocks and
metamorphic rocks that are not being deformed.
As the large crystals grow they need to be accommodated within a fixed
volume of rock. They will tend to rotate to fit into the space created by the
solution of the smaller crystals. If the initial rock has a strong shape-preferred
foliation then this process will reduce the quality of that foliation (Higgins, 1998).
Pure shear, or compaction, is not a very efficient process at producing a
significant fabric (see the above section). An alternative is pressure-solution
compaction (Meurer & Boudreau, 1998). Here, the movement of the crystals
under pure shear is augmented by recrystallisation: plagioclase grains that are
parallel to the applied stress will dissolve more readily than those that are
normal to the stress. This is the same process that is seen in the diagenesis of
sedimentary rocks and occurs because of the excess energy where two grains
are in contact. This process will yield a SPO, but not an LPO. However, an
existing LPO could be preserved. Nicolas and Ildefonse (1996) have proposed
that pressure-solution may also augment simple shear of crystal-rich magmas
(see Section 5.2.2.2)
5.3 Introduction to fabric methodology
A large number of analytical methods have been used to examine the orientation of grains in rocks (Table 5.1). Some methods measure the orientation of
the grain shape (SPO), others quantify the orientation of the crystal lattices
(LPO) and other methods look at a combination of shape and lattice orientation. Some methods examine the orientation of each individual crystal,
5.4 Determination of shape preferred orientations
171
Table 5.1 Summary of grain orientation analytical methods.
Analytical method
SPO or LPO
Individual
or bulk data
Volume or
section
Serial sectioning
X-ray tomography
Sections: optical, orthogonal
stage
Sections: universal stage
Electron backscatter
diffraction
Anisotropy of magnetic
susceptibility (AMS)
Anisotropy of (partial)
anhysteretic remanent
magnetism (AARM –
ApARM)
Neutron diffraction
X-ray diffraction
SPO
SPO
SPO or partial LPO
Individual
Individual
Individual
Volume
Volume
Section
LPO
LPO
Individual
Individual
Volume
Volume
LPO and SPO
Bulk
Volume
LPO and SPO
Bulk
Volume
LPO and SPO
LPO and SPO
Bulk
Bulk
Volume
Volume
whereas others only give a measure of the overall or bulk grain orientations.
Grain orientation is a three-dimensional (volume) property but orthogonal
examination of thin sections only gives orientations in the section; hence
several nonparallel sections must be examined.
5.4 Determination of shape preferred orientations
5.4.1 Direct, three-dimensional analytical methods
The orientation of individual grains can be determined by serial sectioning (see
Section 2.2.1) or X-ray tomography (see Section 2.2.3). Voxel or vector images
can be analysed using the methods mentioned in Section 2.3: each grain is
approximated by a triaxial ellipsoid and these ellipsoids summed to give the
overall fabric ellipsoid. So far, there are no orientation studies yet published
using these techniques. In transparent materials, such as volcanic glasses, the
orientation of crystals can be measured directly with a regular or confocal
microscope (see Section 2.2.1; Manga, 1998).
5.4.2 Section and surface analytical methods
Orientation data can be easily acquired from rock surfaces, slabs and sections,
using a variety of techniques described in Section 2.2. Such images can be
172
Grain orientations: rock fabric
processed manually or automatically to produce classified mineral images (see
Section 2.3). The classified images can be processed to extract overall orientation data using a number of methods described below. Data from orthogonal
sections can be combined to give 3-D orientations.
Imbrication has been quantified by simply counting the imbrication directions of suitable pairs of touching grains (e.g. Ventura et al., 1996). A significant
difference between the numbers of pairs that indicate dextral (clockwise) and
sinistral (anticlockwise) movements is used to define the overall sense of simple
shear. If the flow direction can be established by other means (e.g. geological
structures or palaeosurface orientation) then the grain orientation can be used
to determine the imbrication orientation and hence the direction of flow.
5.4.3 Extraction of orientation parameters
Orientation analytical data need to be summarised before they can be applied
to geological problems. Most work in this subject has been applied to sections,
but can be extended easily to three-dimensional data. A number of different
approaches are possible: for instance individual crystal outlines can be summarised by the inertia tensor method, whereas the intercept method looks at
the orientation distribution of the grain outlines.
5.4.3.1 Inertia tensor method
Any complex grain outline can be reduced to an equivalent ellipse that has the
same area and moment of inertia (Figure 5.4). The orientation ellipse of each
separated grain in a section is calculated by the inertia tensor method and the
results summed for all grains. Clearly, the method can only be used where the
individual grains have been separately defined. The method can also be
extended to 3-D structures. The inertia tensor is calculated from the coordinates of the grain outlines, or from the pixels of a bitmap image (Launeau &
Cruden, 1998). The inertia tensor method has the advantage that the contribution of each grain to the total orientation ellipse is weighted by its area (see
Appendix; Launeau & Cruden, 1998).
5.4.3.2 Intercept (intersection) method
The intercept method is a powerful technique for the numerical analysis of
fabrics (see Appendix; Launeau et al., 1990, Launeau & Robin, 1996). It actually
quantifies the orientation of the boundary of a phase and hence does not
necessarily give the same information as the inertia tensor method. Because it
only examines boundaries, it can be used to analyse any set of objects (crystals,
clasts, voids, lineaments, etc.) that can be identified in the image. It is not
5.4 Determination of shape preferred orientations
173
Rotation
axis
Figure 5.4 The inertia tensor parameter of a crystal outline. The method is
equivalent to the following process: the crystal outline is considered as a thin
sheet of material. The moment of inertia of the crystal outline for an axis in
the plane of the paper can be calculated by summing the inertia contributions
of the pixels enclosed by the outline. For each pixel the inertia depends on the
distance from the rotation axis. The moment of inertia is calculated for
different orientations of the axis and the maximum and minimum values
recorded. These are used to calculate the parameters of an equivalent ellipse.
necessary that the objects are all separated, as is the case for the inertia tensor
method described above. For example, if a slab of granite stained for potassium
feldspar with sodium cobaltinitrite is scanned, then quartz, potassium feldspar
and plagioclase can be readily distinguished from each other. However, most of
the larger crystals will touch, giving a framework. The intercept method can be
used to determine the SPO of this texture. It should be noted that the intercept
method is more commonly used in stereology to count objects or determine the
ratio of surface area to volume (see Section 2.5.2; Russ, 1986).
The method is based on counting intercepts (intersections): a test line is
drawn through the section and an intercept is counted where the object in
question is exited (Figure 5.5a). In practice many test lines are drawn through
the section at regular intervals and at different angles. The numbers of intercepts are counted for each test line and summed for each angle. The intercept
totals for each angle are divided by the total length of the test lines, giving the
intercept density. The inverse of the intercept density is the mean intercept
length. Both data sets can be plotted as rose diagrams. The orientation or
strain ellipse can be derived from the rose of mean intercept lengths.
The star length and star volume methods are intrinsically similar to the
intercept method (Smit et al., 1998). A point is selected randomly within the
phase of interest (Figure 5.5b). The lengths of lines at different angles passing
through this point are compiled. The star length for an angle is the mean length
of the lines at that orientation. The star volume method is similar, but the cube
of the length is summed giving a different weighting for different crystal sizes.
174
Grain orientations: rock fabric
(a) Intercept method
(b) Star length method
Test line
length l
1
2
3
Figure 5.5 (a) The intercept method. A test line of length l is drawn and the
number of times that a crystal is exited (intercepts) is counted, three in this
case. (b) The star length method. A point is randomly placed within the phase
of interest. The lengths of lines radiating from this point at particular angles
are compiled.
The method has also been extended to the analysis of 3-D voxel images
(Ketcham & Ryan, 2004).
5.4.3.3 Grey-scale image analysis methods
Rock fabrics in two dimensions can be deciphered at different scales using
powerful wavelet analysis techniques (Darrozes et al., 1997, Gaillot et al.,
1997, Gaillot et al., 1999). This technique uses 2-D wavelets to give estimates
of the orientation and quality of orientation of mineral grains at different
scales. It is the only generalised method that can do this. It is not necessary to
produce a classified mineral image as the method can be applied to grey-scale
images. Although the method has much promise it has not been used much.
Another image analysis technique that can be applied to grey-scale images is
based on the autocorrelation function (Heilbronner, 1992, Heilbronner, 2002).
This function is used to identify periodic structures in images. It is less generalised and powerful than wavelet-based methods. However, it is better supported by software. A similar method is the directional variogram (Cressie,
1991). It is defined by the variance of the grey-scale pixels at a distance h in a
direction a.
5.5 Determination of lattice preferred orientations
5.5.1 Optical analytical methods: regular and universal stages
Partial information on the lattice orientation of some minerals can be obtained
from a regular thin section and petrographic microscope. For uniaxial
5.5 Determination of lattice preferred orientations
175
minerals, such as quartz and calcite, the optical axes are parallel to the crystallographic axes; hence extinction positions in cross-polarised light can indicate
the orientation of a plane that contains the extraordinary ray or rotational axis
of symmetry.
For biaxial minerals the optical axes are not generally parallel to the crystallographic axes, hence structural features such as twins or cleavage are necessary to define the lattice orientation. Some information on the lattice
orientation of minerals can be readily obtained from the orientation of the
trace of the twin lamellae. This is most commonly applied to plagioclase.
The universal stage is an accessory for a petrographic microscope that
enables a standard thin section to be rotated out of the plane of the microscope
stage. A standard thin section is sandwiched between two glass hemispheres
and the whole assemblage can then be rotated along four or five independent
axes (Emmons, 1964, Muir, 1981). Most universal stages have to be read
manually, but some have been modified by the addition of sensors that transfer
data to a computer. Two or three orthogonal sections must be measured, as the
section can only be rotated up to 458 out of the plane of the stage.
It is relatively easy and quick to measure the orientations of uniaxial
minerals such as quartz and calcite, as only one direction need be determined.
Orthorhombic minerals, such as olivine, are more complex as three directions
must be determined. The most difficult are monoclinic and triclinic minerals,
such as plagioclase. The relationship between the lattice orientations and the
axes of the optical indicatrix is controlled by the composition of the plagioclase. If the composition is known then the lattice orientation can be determined from measurements of optical axes, cleavage planes and twin
composition planes. The method is complex but a computer program can
ease the calculations (Benn & Mainprice, 1989). Normally grains are measured
in three orthogonal sections. However, Ji et al. (1994) suggest that measuring
all twinned grains in any two orthogonal sections is more statistically meaningful and time-saving. However, their method does not always provide a
complete LPO.
5.5.2 Computer-integrated polarisation microscopy
Partial information on the orientation of anisotropic grains in a section can be
obtained from their extinction positions in cross-polarised light. A computercontrolled microscope stage has been developed that can automate this process
(see Section 2.6.4; Heilbronner & Pauli, 1993, Fueten, 1997, Fueten &
Goodchild, 2001). This system can give information on individual grains as
well as the overall fabric ellipsoid of the rock. In general, it can only be applied
176
Grain orientations: rock fabric
to rocks containing a single mineral with an orthorhombic or higher symmetry. Plagioclase, for example, cannot currently be measured using this technique, because of its low symmetry and almost ubiquitous twinning.
5.5.3 Electron backscatter diffraction (EBSD)
Electron backscatter diffraction (EBSD) is a versatile technique that can give
the complete crystallographic orientation of individual mineral grains as small
as 1 mm in polished thin sections and surfaces (see review in Prior et al., 1999).
It can be used with all minerals, even those with low symmetry, such as
plagioclase. It uses a standard scanning electron microscope (SEM) equipped
with a phosphor screen and a light detector. Many systems are completely
automated and can collect data from crystals in an area of 5 mm2.
When an electron beam strikes a surface, electrons will be scattered by the
atoms of the material in the top few mm (see Section 2.5.2.1). This omnidirectional population of electrons is the source for EBSD in an SEM. Electrons are
diffracted by lattice planes in a crystal to produce two cones, with axes normal
to the lattice plane. Under normal conditions for EBSD the opening angle
between the two cones is very small; hence the electrons appear to stream out
along the lattice planes. These electrons are allowed to strike a phosphor
screen where they are converted into light. This light is detected by a CCD
and conveyed to the computer. Many lattice planes in a crystal will diffract
electrons and these bands will interfere constructively. Where a number of
bands intersect, a bright spot will form which corresponds to a zone axis.
Crystallographic directions can be identified from the relative position of the
bands and zones. This process is commonly automated, although there is
always a proportion of patterns that cannot be indexed. Spatial resolution is
about 1 mm in geological specimens, but can be improved by using a field
emission electron gun.
There are always compromises to the optimal configuration for EBSD,
especially as the necessary hardware is commonly added onto an existing
SEM. The sample surface is oriented at an angle of about 708. The phosphor
screen must be as close as possible to the part of the sample illuminated by the
electron beam. It is better to move the sample and not the beam so that the
geometry remains constant.
The importance of correct sample preparation cannot be overstated. EBSD
observes the crystal lattice in the top 1–2 mm of the surface; hence the lattice
must be intact there. Crystal faces and fractures provide a pristine lattice, but
are not very useful for most projects. Mechanical polishing generally damages
the lattice, although some workers have had success with a very fine colloidal
5.6 3-D bulk fabric methods
177
silica (Xie et al., 2003). A number of other treatments can be used: etching
removes the damaged material, but generates unwanted relief. Chemicalmechanical polishing gives a good surface, but some minerals dissolve rapidly
in the fluid used (Prior et al., 1999). Ion-beam milling is another promising
possibility.
Another problem for geological specimens is that they are generally nonconducting and hence surface charging may occur. This effect is very variable
both spatially and temporally. Coating the sample in carbon reduces the
problem, but requires higher beam currents. Most studies do not use coated
samples, but just accept that some measurements will be lost owing to charging. Another possibility is to maintain the sample chamber at a higher
pressure in an environmental SEM (Habesch, 2000). This results in more
electron scattering but less sample charging.
5.6 3-D bulk fabric methods: combined SPO and LPO
A number of methods examine the bulk physical properties of the grains of one
or all phases in a rock. In general such methods record a combination of the
SPO and LPO. They give a single value for the orientation ellipsoid for the
material.
5.6.1 Anisotropy of magnetic susceptibility
Anisotropy of magnetic susceptibility (AMS) is a technique for measuring the
bulk, 3-D magnetic fabric of rocks. It is sensitive, quick, and relatively cheap
and can be applied to a wide range of rock types (Jackson, 1991, Rochette
et al., 1992, Tarling & Hrouda, 1993, Rochette et al., 1999, Martin-Hernandez
et al., 2005). It has been applied extensively in all branches of structural
geology and petrology. If the susceptibility is dominated by magnetite then
the AMS measures only the SPO and spatial distribution of that mineral. If
there are significant contributions from paramagnetic minerals then the AMS
records a mixture of the SPO and LPO. AMS is measured using core samples
25 mm in diameter and 22 mm long. The cores are oriented in the field or
drilled in the laboratory from oriented blocks. Many samples are generally
taken at each outcrop or unit of interest and the results averaged. AMS is
measured in a dedicated instrument.
The total magnetic susceptibility of a sample, K, is the ratio of the intensity
of magnetisation of a material divided by the strength of the inducing magnetic
field. It is not necessarily related to the remanent magnetism, which is the
magnetic field of a sample in the absence of an inducing field. The magnetic
178
Grain orientations: rock fabric
susceptibility is commonly anisotropic and this anisotropy has an orientation:
hence, AMS is second order tensor, Kij and can be represented as an ellipsoid.
Three groups of minerals contribute to the susceptibility (Rochette et al.,
1992). (1) Ferrimagnetic minerals include those that are generally considered
to be ‘magnetic’, that is they have a significant remanent magnetism. They
have a much higher susceptibility than other minerals and will dominate the
susceptibility of a rock when the measurement is carried out under normal
conditions, in a low field. The most well-known ferrimagnetic minerals are
magnetite, ilmenite, pyrrhotite, garnet and hematite. (2) Paramagnetic minerals have lower susceptibilities, but may be present in larger quantities than
ferrimagnetic minerals. Those minerals that are richer in iron will generally
have a higher susceptibility. Important minerals of this group are pyroxenes,
amphiboles, staurolite, cordierite and biotite. (3) Diamagnetic minerals have a
very low negative susceptibility; this means that the induced field has an
opposite polarity to the applied field. These minerals are only important in
the absence of ferrimagnetic or paramagnetic minerals. Prominent diamagnetic minerals are quartz and calcite. The total anisotropy of magnetic susceptibility of a rock is produced by the anisotropy of magnetic susceptibility of the
mineral grains, if they have a LPO. It is also produced by the SPO of minerals
which have a significant total susceptibility, but which are not necessarily
anisotropic (e.g. garnet).
AMS analysis produces a susceptibility ellipsoid whose principal axes
are labelled k1 k2 k3 (or kmax, kint and kmin). The parameters of the
ellipsoid are commonly reduced to orientation, mean susceptibility (km ¼
(k1 þ k2 þ k3)/3), lineation parameter (L ¼ k1/k2), foliation parameter
(F ¼ k2/k3) and degree of anisotropy (P ¼ k1/k3). The shape of the magnetic
anisotropy ellipsoid can be displayed on a ‘Flinn-type’ diagram (Flinn, 1962),
where L is plotted against F.
Most interpretations of AMS data are based on the following assumptions
(Borradaile, 1988, Rochette et al., 1992, Borradaile & Henry, 1997). (1) The
AMS ellipsoid is coaxial to the fabric, with the k3 axis perpendicular to the
foliation and the k1 parallel to the lineation. (2) The shape of the AMS ellipsoid
is related to the rock fabric, in terms of intensity of foliation and lineation
development. (3) AMS measurements are not affected by remanent magnetism. These assumptions must be verified by comparison with independent
mineral fabric measurements using any of the methods described in this
chapter. Where they are correct the magnetic fabric is called ‘normal’. In
some cases it is found that the k1 axis is perpendicular to the foliation and
the k2 axis is parallel to the lineation. This is called a ‘reverse magnetic fabric’.
In other cases k2 is perpendicular to the foliation, which is referred to as an
5.6 3-D bulk fabric methods
179
‘intermediate magnetic fabric’. Finally, the magnetic axes may have no relation to the structure.
In some situations reverse or intermediate magnetic fabrics may have been
identified on structural, rather than mineral fabric criteria. Hence, it is possible
that the magnetic fabrics are in fact normal and the complexity of the structure
has been underestimated (Rochette et al., 1992). Reverse and intermediate
magnetic fabrics can be identified by rock and magnetic fabric measurements
(Geoffroy et al., 2002); however, in these cases the magnetic measurements are
made redundant by the fabric analyses. A more useful way of detecting reverse
and intermediate fabrics is by identification of the magnetic minerals that
produce the susceptibility. A number of minerals, such as single-domain
magnetite and tourmaline, can produce an inverse fabric (Rochette et al.,
1992). The magnetic mineralogy can be identified using optical means, but
commonly many magnetic minerals are present. In these cases other magnetic
methods can be used (see Section 5.6.2).
5.6.2 Anisotropy of (partial) anhysteretic remanent magnetisation
Many rocks have complex histories which are reflected in more than one
fabric. In many situations a weak, late, geologically insignificant magnetic
fabric can dominate over fabrics developed in response to more important geological events. The AMS of a sample is contributed by all ironrich minerals. However, anisotropy of anhysteretic remanent magnetisation
(AARM ¼ ARM ¼ anisotropy of isothermal remanent magnetism) and the
related method anisotropy of partial anhysteretic remanent magnetisation
(ApARM ¼ pARM) respond only to ferrimagnetic minerals, essentially magnetite, hematite and pyrrhotite (Jackson et al., 1988, Jackson, 1991).
Remanent magnetism is the magnetic field produced by a sample in the
absence of an applied field. It is produced by the ferrimagnetic minerals and
can be used in two different ways. The direction of the natural remanent
magnetism of samples (NRM) is used in palaeomagnetic studies to establish
the direction of the magnetic poles with respect to the sample. Similarly, if other
data are known then the NRM can be used to correct for tilting and rotation of
the unit since the fabric was produced (e.g. Palmer & MacDonald, 2002). The
NRM of a sample can be erased by thermal or alternating field demagnetisation
and a new magnetism induced. This magnetic field will be related to the anisotropy of the ferrimagnetic minerals in the rock. AARM examines the total
anisotropy of remanence in a rock, produced by all ferrimagnetic grains. The
strength of remanence is different for each grain and this can be used in ApARM
to separate populations of grains and examine their fabric independently.
180
Grain orientations: rock fabric
The strength of remanence is measured by the coercivity, which is the
intensity of the magnetic field required to reduce the magnetisation of that
material to zero after the magnetisation of the sample has reached saturation.
‘Hard’ materials have a high coercivity and ‘soft’ materials a low coercivity.
The coercivity depends mostly on the nature of the mineral, its size and shape.
The fabrics of grains with different coercivities can be determined by applying
a weak, constant magnetic field at the same time as a stronger alternating field
is allowed to decay (Jackson et al., 1988). Alternatively the sample can be
magnetised completely and then low coercivity fractions removed progressively by alternating field demagnetisation (Trindade et al., 2001). The anisotropy of remanent magnetism within different coercivity intervals is called the
partial anhysteretic remanent magnetisation. It has proved to be particularly
useful for isolating fabrics produced by late secondary alterations (e.g.
Trindade et al., 2001).
5.6.3 Other integrated (bulk) analytical methods
Neutrons are diffracted by atoms and can therefore be used to determine the
lattice parameters and orientation of crystals. Neutron beams cannot be
focused; hence the method uses a broad beam that illuminates many crystals
and gives bulk properties. The weak interactions between neutrons and most
rocks mean that large samples must be used. Hence, the method is attractive
for rocks with large grain sizes. Almost no sample preparation is needed: some
workers shape their samples into 1 cm-diameter spheres (Xie et al., 2003);
others just use irregular blocks (Schafer, 2002).
Pulsed neutron sources enable the use of time-of-flight in addition to
angular position information for analysis (Schafer, 2002, Xie et al., 2003).
Recent developments in software have enabled the method to analyse
low-symmetry minerals such as plagioclase. The method is limited by the
small number of neutron sources and detector arrays that can be used for
these analyses. Sample analysis time is generally long and demand for these
facilities intense, hence few samples can be analysed. Synchrotron X-rays can
also be used and this may be the source of choice in the future (Wenk &
Grigull, 2003).
X-ray diffractometry can also be used to determine the orientation of
mineral grains in a limited class of materials. It is usually only practicable
for minerals with an orthorhombic or higher symmetry. Mixtures of many
minerals may also cause problems, as the diffraction lines cannot always be
separated and identified. The low penetration power of X-rays in most rocks
means that only the surface of a sample is examined. Despite these restrictions
5.7 Extraction of grain orientation data and parameters
181
the low cost of the apparatus, when used on an existing X-ray diffractometer,
can make the method attractive (Rodriguez-Navarro & Romanek, 2002).
5.7 Extraction of grain orientation data and parameters
The methods described above give a variety of different types of data: some
techniques measure the bulk properties of the sample, whereas others give
information on individual crystals in a rock. Individual crystal data are much
richer but may not always be simple to interpret. Hence, individual data are
commonly reduced to an orientation ellipsoid ( second rank tensor), which is
very useful for treating large numbers of samples or comparing data with
methods that measure bulk properties (see below and Mardia & Jupp, 2000).
The orientation ellipsoid should be distinguished from the strain ellipsoid,
which is a measure of the deformation required to produce the orientation
ellipsoid from a population of randomly oriented lines. In a 2-D section the
two are identical, but in 3-D they may differ: the directions of the axes are
identical, but their relative lengths may be different (Gee et al., 2004). If we
have 3-D data, such as EBSD, then we can calculate the orientation ellipsoid
directly from these data. However, many popular methods can only be applied
to sections: here we must integrate data from several non-parallel sections to
get the 3-D ellipsoid (see below).
5.7.1 Individual grain orientation display
Individual crystal data are commonly displayed graphically so that they can be
interpreted in terms of geological processes. Three angular parameters are
needed to express the three-dimensional orientation of each independent
crystallographic (lattice) axis or shape axis of a crystal. This can only be
expressed in a three-dimensional diagram, such as the orientation distribution
function diagram (ODF) (Schmid et al., 1981, Bunge, 1982). Because of the
obvious difficulties of ODF diagrams geologists commonly use projections of
polar diagrams of pole figures. The most popular is the equal-area net
(Schmidt or Lambert net). SPO can be displayed as the directions of major
and minor (longest and shortest) axes of each crystal with respect to specimen
reference directions. LPO can be displayed in a similar way by plotting poles to
lattice planes, crystallographic axes or zone axes with reference to some
geologically relevant orientation. For uniaxial crystals one net gives most of
the useful fabric information, but two nets are needed for the complete
specification of the orientations of all crystals. Three nets may make some
fabric symmetry more evident. In some studies the inverse pole diagram is used
182
Grain orientations: rock fabric
(Randle & Caul, 1996). Here, the crystal axes are the reference frame and the
orientation of the lineation plotted for each crystal.
Two-dimensional orientation data can be readily displayed as we require only
two parameters. The simplest diagram is a conventional histogram of crystal
numbers against angle. Data bins must be chosen with regard to the number of
data points: more detail is possible if more data are available. A variation on
this theme is the rose diagram, which is a polar histogram of the number of
crystals against real angle. Such a diagram is symmetric, because crystal orientations do not have a distinct direction. Such a rose of orientations is a periodic
function that can be analysed by the Fourier method. This can have as many
terms as the number of data points, but for real data it is best to truncate it
(Launeau & Robin, 1996). The size and shape of the data window analysed may
be important if few grains are present. In this case edge effects must be considered; that is grains that lie across the boundary of the analysed area. The
simplest solution is either to increase the area measured or to use a circular
section. However, the latter assumes that the fabric is not strongly developed.
Orientation data can also be displayed without dividing into bins in the
cumulative distribution function against angle diagram. This is a plot of the
number of crystals with an orientation angle less than a certain value against
the angle. It is easy to apply some statistical tests to this function as it increases
monotonically from zero to n, the number of samples. It is also used by Gee
et al. (2004) to combine sectional data into an orientation ellipsoid (see
Section 5.7.3).
The variation in crystal orientation across a surface can be coloured according to a ‘look-up’ table. The table contains various bins with their corresponding colours. The colours are then applied to a map of the surface or volume.
These diagrams are sometimes called AVA diagrams.
5.7.2 Mean orientation from individual data
The statistics of 3-D orientations are complex, but have been well developed
and described (e.g. Mardia & Jupp, 2000). For AMS studies the accepted best
way to calculate mean orientations is that of Hext (1963) and Jelinek (1978).
Such statistical methods are commonly incorporated into the programs that
reduce AMS data.
The statistics of 2-D data are simpler but present a few complexities, hence
are briefly described here (see also review in Capaccioni et al., 1997). We are
usually interested in the mean and degree of dispersion of the orientation data.
cannot be calculated as a simple mean
The mean grain orientation angle, Y,
because the orientation variable is circular, that is 08 ¼ 1808. Note that this
5.7 Extraction of grain orientation data and parameters
183
method is slightly different from that for direction data, which varies from 0 to
3608. The simplest method uses the sum of the direction vectors, with the
angles doubled so that the data vary from 0 to 3608:
X
X
xr ¼
sin 2Yi and yr ¼
cos 2Yi where Yi is the orientation
of the ith outline:
¼ tan1 ðxr =yr Þ=2
Y
Note that tan1 must be corrected for the quadrant. The length of the sum of
the direction vectors divided by the number of measurements can be used as a
test of significance, or departure from random distribution, or can be a
measure of the geological dispersion:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
2 X
1 X
2
R¼
ðcos Yi Þ
sin Yi þ
n
Statistical tables are available to test if the distribution is significant (Swan &
Sandilands, 1995).
Another approach is to use the eigenvalue associated with the first eigenvector of the direction cosine tensor, T (Harvey & Laxton, 1980, Launeau &
Cruden, 1998):
P
P
sinðYi Þ cosðYi Þ 1 sinðYi Þ2
T ¼ P
P
N sinðYi Þ cosðYi Þ
cosðYi Þ2 where N ¼ number of crystals and Y ¼ orientation of crystals. The dispersion
of the orientations is derived from the eigenvalues, e1 and e2. For no alignment
e1 ¼ e2 ¼ 0.5; for perfect alignment e1 ¼ 1.0 and e2 ¼ 0. Several
are
ffiffiffiffiffiffiffiffiffiffiffi
pparameters
used: one is the ratio of the axes of the orientation ellipse ¼ e1 =e2 . For this
measure massive rocks have a value of one and perfectly foliated rocks have a
value of zero. Another is the alignment factor ¼ 2(e1 0.5), (or 100 times this
value). In this case massive rocks have a value of zero and perfectly foliated
rocks have a value of one.
5.7.3 Integration of 2-D section data into a 3-D orientation ellipsoid
Some of the techniques described above give information on sections through a
rock. Data from at least three non-coplanar sections are needed to give a threedimensional orientation ellipsoid that summarises the data (Launeau, 2004).
Ideally, one section should be parallel to the foliation, a second orthogonal to
the foliation and parallel to the lineation and the third orthogonal to the
lineation. However, if the foliation or lineation is weak then this is difficult.
184
Grain orientations: rock fabric
The sectional ellipse method starts with the reduction of the data from each
section to a 2-D orientation ellipse (see above). If simple orientation data are
measured, such as crystal lattice or shape orientations, then the orientation of
the major axis of the ellipse is clear but the absolute size is unknown: instead,
we just know the ratio of the major and minor axes. Robin (2002) has developed a robust method that gives the best-fit ellipsoid from three or more
sections and a quality-of-fit parameter (see Appendix). However, Gee et al.
(2004) consider that this method does not give an accurate ellipsoid for strong
fabrics, especially where the sections are not cut accurately along the significant planes of the ellipsoid.
The approach of Gee et al. (2004) is stochastic. They initially reduce the
orientation data in each section to the cumulative distribution function (CDF;
see above). The contribution of each measurement to the CDF can be weighted
by the crystal length or area. They then take a population of randomly
oriented vectors (representing the crystals), apply a strain tensor to their
orientations (equivalent to deformation) and calculate a CDF for each section.
The resulting CDFs are compared to the data CDFs for all the sections and the
total misfit calculated. The parameters of the strain tensor are then varied to
minimise the misfit between the CDFs. The orientation ellipsoid can be
calculated from the strain tensor that gives the smallest misfit (see
Appendix). The uncertainties in the orientation ellipsoid parameters were
also calculated. The method was tested on two samples and appears to give
an orientation ellipsoid that conforms closely to that calculated from individual crystal measurements by EBSD. It was much better than the sectional
ellipse method for a sample with a strong fabric.
5.8 Typical applications
Applications were chosen because they illustrate what people are doing, or
because they show interesting approaches. AMS has been used for a large
number of studies in all branches of petrology. Recent collections include a
special issue of Tectonophysics (307, 1–234), a book (Martin-Hernandez et al.,
2005) and a review article (Borradaile & Henry, 1997).
5.8.1 Flow in lavas
Although movements of the exterior of lava flows can be observed easily, it is
generally necessary to determine the directions of movements within the flow,
so that the exact mechanism of flow may be established. A number of models
are possible: plug, Bingham or Newtonian flow, similar to flow in an enclosed
5.8 Typical applications
185
conduit; glacier type flow; or rolling of the lava front beneath the flow. Some
of these possibilities have been explored for various lava compositions.
5.8.1.1 Mafic and intermediate lavas
A short, crystal-rich dacite lava flow on Monte Porri, Salina, Italy was
investigated by Ventura et al. (1996). A single detailed section through the
central 2.2 m-thick massive part of the flow was examined with 19 samples.
The authors measured tabular plagioclase crystals with axial ratios of 4 and
quantified their orientations and sense of imbrication from oriented thin
sections. Crystals with lengths of 0.2–1.0 mm had orientations that varied
consistently through the flow section: at the base and top the crystals were
aligned parallel to the base and top with the strongest fabric. Towards the
interior, the fabric became more inclined, up to 408, and weaker. In the central
part of the flow the fabric was not evident. Imbrication directions agreed
statistically with the flow sense indicated by the crystal orientations, but
30%–40% of individual data were in the opposite direction. Crystals from
0.04 to 0.2 mm long were randomly oriented and appear to have grown in place
after flow finished. The authors conclude that the textures accord with a plug
flow model between two rigid walls. This is a bit surprising as the upper part of
the flow is not confined. Maybe the scoriaceous breccia zone acts as a brake on
movements within the flow or perhaps magma continued to move in the flow
after the upper surface had solidified.
Many lava flows and plutonic rocks contain ellipsoid bodies of contrasting
composition, which can also be used to determine flow directions and mechanisms. These enclaves result from mixing of magmas and it is important to
determine if the enclaves have deformed internally during magma movement
or if they have merely rotated as rigid bodies. A number of factors intervene,
but the most important are probably the viscosity contrast with the host
magma and the yield strength of the enclave and host magma (Ventura,
2001). The Monte Porri dacite lava flow was used as an example (Ventura,
2001) because crystal orientations had already been used to determine the flow
regime (Ventura et al., 1996). Enclave orientation and shape were measured in
vertical sections near the base of the flow and used to establish the proportion
of pure shear (compaction) to simple shear. Near the vent flow was hyperbolic,
reflecting a mixture of spreading of the flow and gliding (Figure 5.6). The
component of simple shear increased with distance along the flow indicating
an increase in the gliding component.
Egmont volcano is a fairly typical stratovolcano covered with andesite
flows. Most are thin, and have commonly almost drained away, but there
are a number of much thicker flows. The bases of two flows were examined by
186
Grain orientations: rock fabric
90
Pure shear flow
(compaction)
80
70
Θ (°)
60
50
Hyperbolic
flow
40
30
20
10
0
100
Simple shear
flow
300
500
700
900
distance from the vent (m)
1100
1300
Figure 5.6 Nature of shear in the Monte Porri dacite lava derived from
measurements of the orientation and shape of ellipsoid enclaves (Ventura,
2001). The angle Y is the angle between the streamlines and the base of the
flow. It is 908 for pure shear (compaction), 458 for hyperbolic shear and 08 for
simple shear.
Higgins (1996a): the Kokowai Valley flow is about 10 m thick, and
Humphries’ Castle is over 100 m thick. The SPO of plagioclase was determined
optically using thin sections (Figure 5.7). In both cases the preferred orientation direction showed a consistent pattern: near the base the crystals are
oriented parallel to the slope of the flow, but above a height of 1–2 m the
preferred orientation dips at about 308 in a downstream direction. This is
opposite to the direction indicated in the Monte Porri dacite flow (Ventura
et al., 1996). The significance of the intermediate sample from the Kokowai
Valley flow that dips upstream at about 208 is not clear. The quality of
plagioclase orientation does not appear to correlate strongly with sample
position. These orientations suggest that the flow advanced with a rolling
motion: crystals were aligned during flow further upstream and then rolled
over the stagnant zone close to the base of the flow. However, more data is
needed to confirm this.
Traditionally, only the surface topology of lava was used to classify the flow
as a’a, pahoehoe or toothpaste. However, the differences between these types
are more fundamental and continue into the interior of the flows: differences in
CSDs have been noted in Section 3.4.1.2. The transition from pahoehoe to
toothpaste to a’a is related to both the melt viscosity and the strain rate.
5.8 Typical applications
187
Humphries' Castle
Kokowai Valley
Flow direction
height above flow base (m)
2.0 m
1.0 m
Base of flow
Figure 5.7 Orientations of plagioclase crystals in basaltic andesites from
Egmont Volcano (Mt Taranaki), New Zealand (Higgins, 1996a). Two
different flows were sampled, each at three heights. SPOs are displayed as
rose diagrams.
Canon-Tapia et al. (1997) investigated these differences in a series of basalt
flows from Hawaii using AMS. They surmised that the magnetic characteristics of the lavas can be explained by the distribution of the magnetic particles
in the lava and post-emplacement crystallisation. However, the orientation
and degree of anisotropy of magnetic susceptibility is defined by shear of the
magma during consolidation. Although some authors have suggested that k2 is
parallel to the flow direction, Canon-Tapia et al. (1997) consider that here it is
k1. In all the flows the shear rate was found to be most intense at the base of the
flow, as is to be expected. However a’a flows were in general more sheared than
pahoehoe flows. The directions of AMS within the flows seem to be distributed
more regularly in a’a as compared to the pahoehoe flows. This suggests that
a’a continues to flow until it is stopped by high viscosity. In contrast, pahoehoe
flows appear to cease flowing when the magma is still fluid. This enables late
188
Grain orientations: rock fabric
flow and/or endogenous growth after the flow has stopped and dispersed AMS
directions.
5.8.1.2 Felsic lavas
Some microlites are prismatic and hence provide almost ideal markers for
determining the amount of simple shear and pure shear (compaction) in lava
flows and domes (Manga, 1998). Some glassy felsic volcanic rocks are transparent enough that the 3-D orientation of microlites can be measured directly.
Obsidian Dome, USA, is a rhyolite flow about 3 km is diameter: orientations
of pyroxene microlites show that the flow developed by radial flow followed by
flattening: pure shear increases from 0.3 at the vent to 1.1 at the flow front
(Castro et al., 2003).
AMS measurements were also made on the same rocks (Canon-Tapia &
Castro, 2004). Although the bulk susceptibility of the rocks is low the degree of
magnetic anisotropy is extremely high. Directions of the maximum axis
accorded with the lineation direction of the microlites and showed the local
direction of flow. The authors caution that their results may not be applicable
elsewhere if the susceptibility is not caused by the same minerals with similar
prismatic forms.
5.8.2 Flow in dykes
The textural fabric of dykes is of immense importance as it can be used to
determine the direction of flow of the magma (Blanchard et al., 1979). Hence,
it is possible to locate the position of the original magma chambers, which can
in turn tell us about regional geodynamics. Many studies have used AMS to
determine flow directions (e.g. Ernst & Baragar, 1992), but there are significant problems in the interpretation: it is not always clear how the magnetic
vectors correspond to the magmatic deformation field. In some cases k2
appears to be parallel to flow direction, whereas in others it is k1 (Geoffroy
et al., 2002). There are a number of ways in which this uncertainty can be
resolved.
Optical methods can be used to determine the flow direction of magma in
phyric dykes. Oriented thin sections are cut from lateral positions on either
side of the dyke and the imbrication of magmatic foliation can be used to
determine the flow direction (Figure 5.8). In one of the earliest studies a section
was cut parallel to the dyke walls and the orientation of tabular plagioclase
crystals determined using a universal stage (Shelley, 1985). Such LPO measurements are very tedious hence more recent studies tend to examine only the
SPO of phenocrysts. This is readily done directly from a thin section using the
5.8 Typical applications
189
Figure 5.8 Flow of magma with crystals (or enclaves) in a dyke. Initially
randomly oriented crystals develop a strong fabric near the dyke walls that
angles into the flow centre towards the direction of flow. This imbrication is
mirrored on the other side of the dyke. Fabric is not developed in the centre of
the dyke.
intercept method (Geoffroy et al., 2002). In one study of a composite dyke the
directions of emplacement of the three separate phases could be identified
(Wada, 1992). In some studies the flow direction was determined by optical
methods and used to establish the axis of the corresponding AMS ellipsoid.
The AMS could then be applied to a much larger group of samples (e.g. Poland
et al., 2004). Another possibility is to look at the imbrication of AMS ellipsoids
from samples taken on either side of the dyke, a process analogous to the
optical determination (Geoffroy et al., 2002).
Mafic dyke swarms can cover huge areas and individual dykes can sometimes be traced for hundreds of kilometres. Ernst and Baragar (1992) examined
the McKenzie dyke swarm, which radiates from a point in northern Canada
and extends southward for 2000 km. Such dyke swarms are commonly
associated with mantle plumes and it is of interest to determine the original
area of magma production. Ernst and Baragar (1992) showed that close to the
focus of the swarm the flow direction has an important vertical component,
but this becomes horizontal away from the focus. This shows that mafic
magmas can be propagated over vast distances in the crust.
Magma flow directions in a dacite dyke on a more local scale were examined
by Poland et al. (2004). Two dykes 3 km long were investigated using AMS
with verification of the flow directions from measurements of crystal orientations. The flow directions in the dykes were simple and linear in their deeper
parts, but broke up into en echelon segments close to the surface (Figure 5.9).
190
~1 km
1
2
Er
os
io
na
l
su
rfa
ce
Grain orientations: rock fabric
Continuous dyke
Figure 5.9 Emplacement mechanism of a dacite dyke derived from thin
section analysis and AMS measurements (Poland et al., 2004).
The dominant flow direction in the deeper parts of the dyke was horizontal.
Flow in en echelon segments is generally assumed to be parallel to the axis of
segment rotation. However, this was not found to be generally the case here:
again dyke flow was largely horizontal, either parallel with the overall direction of flow, or into the tips of the segments in the opposite direction.
Although many dykes form by simple extension normal to the walls, dykes
are a zone of weakness and hence can easily accommodate lateral displacement
of the walls, as in a fault. In narrow dykes such displacement can commonly be
seen in the field, but it is not so evident for larger dykes. However, displacement can be deduced from AMS measurements of fabric obliquity near the
walls of the dykes. In simple dykes fabric obliquities are opposed on either side
of the dyke (Figure 5.8). If there has been displacement of the walls during
solidification then the obliquity on one side may be inversed, making it parallel
to that on the opposite side. Scott and Benn (2001) examined dykes radial to
the Sudbury impact structure. Their AMS measurement revealed a steep
foliation that was counterclockwise oblique on both sides of the dyke. They
proposed that the dyke acted as a magma-filled transfer fault during collapse
of the inner ring of the central peak, all formed just after the impact of the
meteorite.
The magnetic fabrics of altered and metamorphosed dykes may be complex
and difficult to understand. Borradaile and Gauthier (2003) examined dykes
from the Troodos ophiolite complex, Cyprus, which had been altered by
hydrothermal fluids. Original magnetite and titanomagnetite were partly
replaced by the more oxidised mineral maghemite. They found that maghemite
was a significant component of the bulk susceptibility, but that AARM
indicated that this secondary magnetic fabric had a low anisotropy. Hence,
AMS measurements corresponded to magmatic flow directions. Clearly, this is
not always the case and AARM determination are useful to decode anomalous
magnetic fabric directions.
191
5.8 Typical applications
5.8.3 Magmatic deformation in plutonic rocks
5.8.3.1 Mafic rocks
Mineral layering and foliation (lamination) is widely observed in mafic plutonic rocks but most studies have been qualitative, rather than quantitative
(Wager & Brown, 1968, Naslund & McBirney, 1996). The upper part of the
Sept Iles intrusive suite contains anorthosite some of which is very well foliated
(Higgins, 1991). The shape and SPO of plagioclase in a foliated sample was
determined using orthogonal thin sections. Higgins (1991) showed that a
foliation of such quality could only be produced by simple shear whilst there
was still silicate liquid between the grains. The lack of an observable lineation
may reflect the shape of the tablets: they are almost equant in the plane of the
largest surface (1:5:7).
The Kiglapait mafic intrusion, Labrador, Canada, has been well studied as
it is undeformed, relatively small and well exposed (Morse, 1969). The lower
zone is troctolite and the upper zone is gabbro. Plagioclase and olivine SPOs
have been determined both for a section through the intrusion and a detailed
study of a single layer (Higgins, 2002b). Single samples were orientated vertically directed towards the centre of the intrusion. SPOs were measured optically and reduced using the inertia tensor method.
Plagioclase SPOs are significant throughout the section, but are strongest in
the lower quarter of the intrusion (Figure 5.10). Simple compaction is not able
(a)
100
(b)
solidification (%)
80
60
40
20
Layered sample
0
0.6
0.7
0.8
stronger
weaker
ratio of orientation ellipse
characteristic length (mm)
1.5
22
44
11
1.0
60
10C
86
70
10A
0.5
Pr
es
91
co
su
m
pa
98 82
10B
10D
re
so
lu
ct
io
n
tio
n
0
0.6
0.7
0.8
stronger
weaker
ratio of orientation ellipse
Figure 5.10 Plagioclase SPO in the Kiglapait intrusion, Canada (Higgins,
2002b). The ratio of the orientation ellipse indicates the quality of the
foliation: 1 ¼ massive; 0 ¼ perfect foliation.
192
Grain orientations: rock fabric
to produce significant foliation (Higgins, 1991): hence the strength of the SPO
must reflect the vigour and stability of the flow within the chamber. Welllayered and foliated rocks near the base of the intrusion may represent a period
of stagnation interspersed by the flow of crystal laden turbidity currents. The
more strongly foliated samples higher up indicate a period of stable convective
flow, which diminished towards the upper part of the intrusion. Olivine also
has a significant SPO in all samples, which is not surprising as olivine is also a
liquidus phase (Morse, 1979).
As was mentioned before compaction alone cannot produce a significant
SPO, unless the crystals change shape by solution and growth (Hunter, 1996,
Meurer & Boudreau, 1998). Such a process is related to both pressure solution
and coarsening. Hence, we would expect to see a correlation between the
strength of the foliation and the characteristic length. Such is observed for
two samples from Kiglapait, suggesting that this process is either only locally
important or only locally preserved.
The Tellnes deposit, Norway, is a large lens of ilmenite-rich norite within a
long jotunite dyke. Measurements of the petrofabric by AMS and image
analysis have been used to determine the mode of formation of the deposit
(Figure 5.11; Diot et al., 2003). Partial anhysteretic remanent magnetism was
used to show that the AMS was produced by coarse magnetite grains. The
significance of magnetic fabrics is not always clear, hence the fabric of some
A B
D
?
100
200
0m
100
0m
200
100
0m
0
500 m
Tellnes Ilmenite Deposit Norite
Åna Sira Anorthosite
Flow lines
0
0m
J
G
200
?
I
E
200
100
H
500
F
?
G
Scale (m)
C
H Horsetail splays
100
0m
?
N
20
T
B
200
DI O
F
A
03
C D E
I
10
0
0
m
?
100
0m
J
?
?
100
0m
??
100
0m
?
10
0
0m
?
Figure 5.11 Flow directions of magma in the Tellnes ilmenite deposit from
AMS measurements (Diot et al., 2003).
5.8 Typical applications
193
samples were also determined using optical methods. Three orthogonal thin
sections were cut and the fabric produced by both ilmenite and magnetite was
quantified using the intercept method. In general the fabric orientation determined from AMS and image analysis agreed within 308, despite the fact that
AMS looks at magnetite whereas the optical methods mostly reflect ilmenite
distributions. The fabric distributions suggest that the deposit was formed by
multiple injections of crystal-rich mushes and that the fabric was developed by
magmatic deformation during emplacement.
The lower part of the Bushveld complex, South Africa, is comprised of rocks
rich in orthopyroxene. Boorman et al. (2004) studied the alignment of these
prismatic crystals as part of a comprehensive textural study. They found that
there was a good correlation between the aspect ratio of the crystals and their
degree of alignment. However, there was no correlation between the degrees of
alignment normal and parallel to the foliation, indicating that there was no
lineation. If the foliation was produced by simple shear (flow) then such
prismatic crystals should have a significant lineation, unlike the tabular plagioclase crystals discussed above. They conclude that the mineral alignment was
produced by pure shear (compaction) accompanied by coarsening. This process is more efficient in rocks dominated by orthopyroxene as the presence of
other phases will pin the grain boundaries. It is also similar to the pressure
solution compaction proposed by Meurer and Boudreau (1998) and seen
sporadically in the Kiglapait intrusion (Higgins, 2002b).
The Oman ophiolite has been the subject of numerous studies as it is well
exposed and relatively unaltered. Yaouancq and MacLeod (2000) studied the
fabric of the gabbros using field measurements, AMS and universal stage
methods. They found that the AMS only followed the megascopic foliation
or LPO in about 2/3 of the foliated gabbro samples. In these rocks the AMS is
controlled by secondary magnetite produced by alteration of olivine along
fractures. In many samples these fracture planes follow the overall fabric.
However, there is no correspondence between the AMS and plagioclase LPO
ellipsoids. Gabbros from the top of the section show no link between AMS and
overall rock fabric: here AMS is produced by primary interstitial magnetite
grains whose distribution is not controlled by the rock fabric. Hence, application of AMS on these types of rock is extremely problematical and it is best
to quantify rock fabric using measurements on mineral orientations, such as
plagioclase.
Compaction of cumulate rocks is thought to be an important process
during solidification, but has rarely been quantified. The recent discovery of
plagioclase crystal chains in the Holyoke basalt flow and other units has
provided a structure that can be quantified and used to determine the degree
194
Grain orientations: rock fabric
of compaction (Philpotts et al., 1998, Philpotts et al., 1999). Grey et al. (2003)
examined the texture of a sample of gabbro from the Palisades sill, New Jersey,
using four different techniques. They first traced the plagioclase crystal chains
from thin section photographs. From these line networks they determined the
anisotropy using the intercept method (see Section 5.4.3.2). They also determined the anisotropy of the chains from the orientation and size of the links
between adjacent crystals and the geometry of the interstitial regions between
the chains. All these methods produced a similar measure of anisotropy which
was interpreted as a result of compaction of 8%. In addition the anisotropy
plunges down a dip, suggesting that the sill was emplaced in its present
orientation and that there has been minor deformation of the mush in this
direction, consistent with the compaction.
The anisotropy was also determined independently of the plagioclase chains
directly from grey-scale thin section photographs. Images of the silicate and
opaque minerals were analysed using the directional variogram method. The
silicate minerals yielded a similar direction and shape of anisotropy to the data
from the plagioclase chains. However, the opaque minerals gave a very different direction of anisotropy, orthogonal to that of the silicates. This was similar
to that provided by AMS analysis of the same material. This suggested that the
opaque minerals crystallised late, perhaps along cracks that opened up during
compaction.
5.8.3.2 Granitoid intrusions
Granitoids are a major igneous component of the middle and upper crust
and were commonly emplaced during deformation events. Numerous studies
of the fabric of granitoid intrusions show that it is remarkably homogeneous, or varies smoothly through an intrusion and hence it records the
deformation that the magma underwent during solidification (Bouchez,
1997, Bouchez, 2000). Some authors consider that the deformation was
produced by flow in the magma chamber during emplacement, whereas
others have shown that the fabrics record regional strain during the last
stages of solidification.
The fabric of granitoids has been extensively examined using AMS, sometimes supplemented by mineral orientation studies (Bouchez, 1997, Bouchez,
2000). If magnetite is present, then it will completely dominate the overall
susceptibility of the rock. Gregoire et al. (1998) examined a magnetite bearing
quartz-syenite and found a good correlation between the magnetite shape
fabric and the AMS ellipsoid. This contrasts with more mafic rocks, in
which the relationship has not always been so simple (e.g. Yaouancq &
MacLeod, 2000). In the absence of magnetite the AMS is dominated by the
5.8 Typical applications
195
shape fabric of paramagnetic minerals, such as biotite and amphibole. If these
minerals were present when the rock was deformed, then the AMS and mineral
fabrics will be identical.
The Mono Creek pluton is a good example of a magnetite dominated system
(Bouchez, 2000). It was emplaced into earlier components of the Sierra
Nevada Batholith, California, during a period of regional dextral transgression. The AMS does not follow the border of the intrusion, as would be
expected if fabric were due to convection or other internal processes (SaintBlanquat & Tikoff, 1997). Instead, the AMS defines a broad sigmoidal shape
that accords with emplacement during tectonism. Towards the end of solidification the deformation was localised and became more intense, producing a
stronger AMS fabric. Deformation continued after solidification with the
production of a shear zone.
In many granitoids rock fabric, as expressed by AMS, clearly indicates a
sense of deformation, but this deformation may be a late event, unrelated to
the main phase of the emplacement of the intrusion. Cruden and Launeau
(1994) have investigated the Lebel syenite intrusion, Ontario. AMS fabrics,
defined by magnetite, indicate that the intrusion has a sheet-like form. They
propose a model in which the magma rose up an E–W fault and was then
injected horizontally as a tongue of magma into the host rocks. Launeau and
Cruden (1998) extended this study by the analysis of mineral fabrics using the
intercept method. They found that the mineral fabrics were generally parallel
to the AMS ellipsoids: however, in some samples bimodal mineral fabrics were
attributed to overprinting or complex movements of grains with different
shapes. The experiments of Iezzi and Ventura (2002) confirm that minerals
with different shapes can behave differently during magmatic deformation.
Crystal size distributions of pyroxene and magnetite were straight and this was
interpreted to mean that these minerals crystallised over a protracted interval.
Hence, the AMS was mostly produced during the main phase of crystallisation
of the body.
The fabric of some granitoids has been investigated using optical methods at
the same time as other aspects of the texture were quantified. The Halle
Volcanic Complex, Germany, is a porphyritic rhyolite intrusion with abundant quartz, orthoclase and quartz phenocrysts (Mock et al., 2003). Optical
measurement of sections, some of them orthogonal, consistently showed significant preferred orientations at 458 and 1358. This was interpreted as conjugate zones, developed during crystal settling. However, the consistent
direction for both vertical and horizontal sections, and the similar, weak
orientations, suggests that it may just be an artefact of the measurement
process.
196
Grain orientations: rock fabric
5.8.3.3 Engineering studies
The mechanical properties of rocks are of fundamental importance to engineers and there have been a number of attempts to link this to rock texture.
Early studies used a single parameter to summarise the effects of texture on
rock properties (Howarth & Rowlands, 1986). The ‘texture coefficient’ is
derived from the orientation and shape of crystals and porosity of the rock,
but does not seem to be a good predictor of rock properties (Ersoy & Waller,
1995). Akesson et al. (2003) examined the effects of rock fabric and texture on
the mechanical properties of rocks. They used the intercept method to quantify
rock foliation and found that there was a good correlation between the degree
of foliation and the aspect ratio of the fragments. In a previous study of
isotropic rock the same group had found a simple relationship between the
area-normalised grain perimeters and rock fragility as expressed as the Los
Angeles (LA) index (Akesson et al., 2001). The foliation index that they
developed can be used to correct the normalised perimeter to that of isotropic
rocks, and hence extend the use of this relationship.
5.8.4 Solid-state deformation
It has been known for some time that the upper mantle is seismically anisotropic, which is caused by LPO of olivine and orthopyroxene. Hence, the
tectonics of the mantle can be determined by remote measurements of its
fabric. Ben Ismail and Mainprice (1998) have investigated the nature of the
link between fabric and seismic anisotropy using laboratory measurements of
fabric in mantle samples. They examined 110 samples from a wide variety of
settings and with a range of textures. They classified the samples into four
groups on the basis of the orientation of their olivine lattices with respect to
lineation and foliation directions. In all cases the [100] direction was parallel to
the lineation and this had the greatest influence on the P-wave anisotropy. The
[010] and [001] directions were much more variable and provided the key to the
fabric classification. It was found that only the [100] and [001] directions
influenced the S-wave anisotropy.
Garnet is a common mineral in metamorphic rocks, but its isometric symmetry hides textural detail in optical studies. However, garnet is not structurally isotropic and hence measurement of lattice orientations can potentially
reveal petrogenetic processes. Prior et al. (2002) used EBSD to examine two
garnet grains from two different rocks: the Glenelg was a deformed grain and
the Alpine was a porphyroblast (Figure 5.12). Each grain comprised many
sub-grains (defined as an angular difference >28) and they compiled the
197
5.8 Typical applications
Glenelg garnet
Alpine garnet
0.09
0.055
0.08
tic
al
eo
re
0.03
0.02
m
do
an
0.03
lR
0.04
0.035
ore
tica
Ra
n
do
m
relative frequency
0.06
983 Random Pairs
0.04
0.025
0.02
Th
e
984 Random Pairs
0.05
118 Neighbour Pairs
0.045
0.015
Th
relative frequency
0.05
100 Neighbour Pairs
0.07
0.01
0.01
0.005
0
0
0
10
20
30
40
50
misorientation angle (°)
60
0
10
20
30
40
50
misorientation angle (°)
60
Figure 5.12 Orientation of garnet sub-grains by EBSD (Prior et al., 2002).
The sub-grains in the Glenelg garnet were produced by solid-state deformation,
whereas those in the Alpine garnet are the result of coalescence of
independently nucleated grains, followed by rotation.
angular difference (misorientation) between adjoining and randomly selected
sub-grains. These distributions were compared with that expected if the subgrains had random orientations. In both cases adjacent grains had lower
misorientations that randomly selected grains. In the deformed garnet the
sub-grains were internally distorted and the misorientations lay along small
circles around rational crystallographic axes. Hence, this the subgrains likely
formed by deformation of an original large grain. In the porphyroblast subgrains are undistorted and misorientations have random orientations. This
grain formed by coalescence of neighbouring grains. However, misorientations are lower than expected for such a process and it must have been
accompanied by sub-grain rotation (Spiess et al., 2001). This would serve to
minimise the total surface energy of the grain, as do grain-boundary migration
(coarsening) and adjustments to the dihedral angles of the crystals (see
Sections 3.2.4 and 4.2.3).
5.8.5 Fabric development in pyroclastic rocks
The products of large explosive eruptions can cover vast areas with readily
identifiable pyroclastic deposits. However, the vent area of these deposits is
198
Grain orientations: rock fabric
not always easy to identify, as it may have been removed or filled in by later
eruptions. Image analysis has also been used to establish the orientation of the
fabric in tuffs (Capaccioni & Sarocchi, 1996). Blocks were imaged and the
clasts electronically separated. The fabric parameters were derived from the
individual grain outlines. However, sedimentological structures may not be
strong enough to be identified and cannot always be applied, for instance in
drill core. In some areas AMS has been used to identify the flow directions of
such tuffs.
AMS has an orientation, but not a direction. Hence, if the major axis of the
AMS ellipsoid (k1) follows the flow lines of the tuff we can get two possible
directions. Luckily, in many tuffs the AMS ellipsoid dips up-flow at angles of
10–308 (Knight et al., 1986, Palmer & MacDonald, 1999). Studies of recent
undeformed tuffs have shown that the method can be used to identify vent
regions. The degree of anisotropy does not correlate with density or degree of
welding, indicating that the AMS was produced as a consequence of flow and
not afterwards (Palmer & MacDonald, 2002).
Notes
1. The term fabric is used here for the orientation of grains in a rock. The equivalent term in
materials science is texture or microstructure.
6
Grain spatial distributions and relations
6.1 Introduction
Grains in many rocks are not distributed randomly in space, but organised
into clusters, layers and chains (Figure 6.1). Such spatial patterns can be
defined in a number of ways: by the presence of grains as well as by their
size, shape, orientation and mineral associations. Such non-random distributions of grains, termed patterns or packings, have been observed in metamorphic, igneous and sedimentary rocks (Rogers et al., 1994, Jerram et al.,
2003). Despite the fact that quantitative investigation of such structures in a
rock was started in 1966 (Kretz, 1966a), there have been relatively few quantitative studies since that time. Many terms have been used to describe nonuniform grain distributions: in igneous rocks grain clusters are referred to as
clumps, clots or glomerocrysts, whereas in metamorphic petrology other
textural terms are used. Another important textural aspect of a rock is the
spatial association of minerals, or lack thereof. This may reflect nucleation,
growth or mineral breakdown processes, but has been much less studied than
the actual crystal positions.
Patterns can be imposed on a rock during its formation by externally
varying conditions or can develop spontaneously. For instance, during deposition of sediments variations in the flow regime or sediment source can produce
layers; these are clearly imposed externally. Some layering in igneous rocks
may result from similar sedimentary processes (Naslund & McBirney, 1996).
Layering in metamorphic rocks can reflect layering of a sedimentary protolith,
and deformation of dykes and enclaves. Layers or other structures are referred
to as templates (Ortoleva, 1994). In some homogeneous materials patterns can
arise spontaneously, or existing templates can be amplified and modified.
These effects are called ‘geochemical self-organisation’ (Ortoleva, 1994).
A classic example of this effect is Liesegang layering. Hence, grain clusters,
199
200
Grain spatial distributions and relations
Ordered
Random
Layered
Clustered
Chains/Foams
Broken layers
Figure 6.1 Schematic grain spatial distributions in two dimensions. Such
patterns may be defined by the presence of grains of a phase, their orientation
and mineral associations or by differences in grain size.
chains and layers appear to be common phenomena, with varying origins, and
should yield petrogenetic information.
6.2 Brief review of theory
6.2.1 Nucleation of crystals
Nucleation occurs when it is energetically favourable to create a new phase, or
more crystals of an existing phase (see Section 3.2). If nucleation occurs
spontaneously within a phase it is termed homogeneous nucleation. The position of the nuclei will be random, provided that the temperature and composition of the liquid are constant. If, however, a region of the liquid is depleted
in the crystal components, for example by crystallisation of an adjacent crystal,
then nucleation may be suppressed there. This may produce megascopic
patterns by geochemical self-organisation (Ortoleva, 1994).
Nucleation can also occur on existing phases, termed heterogeneous nucleation. The position of the nuclei of the new crystals will obviously depend on the
6.2 Brief review of theory
201
location of the host materials. It is unlikely to be random, unless the host
grains are very small and initially randomly distributed. It is more likely that
the nuclei will be clustered. If the crystals do not move significantly during
solidification, this may be visible in the final rock. It is debatable if any magma
is ever completely free of crystals, hence heterogeneous nucleation may be the
normal situation.
Minerals can also form by the breakdown of other solid phases. If two
minerals are formed by such reactions in the solid state then they will be
spatially associated.
6.2.2 Geochemical self-organisation
Many periodic geological structures are thought to originate by variations in
externally controlled parameters. For example, sedimentary layering may
form by variations in the mineralogy and/or size of the particles being deposited. The variation in externally controlled parameters is considered to be a
‘template’ which forces the structure on the rocks. However, patterns can also
arise spontaneously without the intervention of a template: this vast subject is
termed geochemical self-organisation (Ortoleva, 1994, Jamtveit & Meakin,
1999). A number of patterned structures in geology may be produced by this
effect (Ortoleva, 1994):
*
*
*
*
*
*
layers and spots in igneous rocks
metamorphic differentiation and layering
oscillatory zoning in crystals
Liesegang banding and layering
diagenetic patterns
geodes, concretions, agates and orbicules
Geochemical self-organisation may seem to violate the second law of thermodynamics: that all systems evolve spontaneously towards a state of increased
disorder. However, although geochemical self-organisation is clearly a local
increase in order, it must be compensated by a greater decrease in order of the
whole system.
Self-organised patterns originate by the dissipation of free energy and are
hence called dissipative structures. The initial unordered system is subject to
random variations of thermal origin. This ‘noise’ can be decomposed into
periodic fluctuations of all possible symmetries, in the same way that any wave
can be Fourier transformed into an infinite number of harmonic waves. When
the system is close to equilibrium none of these patterns dominates and the
system appears completely unordered. However, when the system evolves
202
Grain spatial distributions and relations
away from its equilibrium state certain patterns may be amplified by positive
feedback loops and an ordering developed. If the structure is observable and
stable it is termed a dissipative structure. Such patterns may be comprised of
layers, spots, spirals, etc. The patterns form by reaction–transport: this
involves solution of some grains, transport of material and growth of other
grains. Some patterns are only maintained by a continuous supply of free
energy; however, the occurrence of dissipative structures in old rocks indicates
that they can be frozen in place and observed long after the activity of the
system has ceased. It should be emphasised that dissipative structures can form
only in systems that are maintained away from equilibrium.
A classic example of patterning is Liesegang banding. In the simplest
situation this consists of bands richer in a phase in a two-dimensional
system. Three-dimensional Liesegang layers also occur, but have been little
studied (Boudreau, 1995). Two different origins have been proposed
(Chacron & L’Heureux, 1999). In the pre-nucleation model, nucleation and
growth of a crystal depletes the surrounding area in that component, inhibiting nucleation of further crystals. As the reaction front moves away, higher
concentrations are encountered and nucleation is enabled again, starting
the production of a new layer. However, Liesegang bands can also form
after nucleation from a homogeneous material by competitive particle
growth and coarsening (see Section 3.2.4). If a small region has a local
increase in crystal size, then the concentration of the component in the liquid
will decrease. As material diffuses in from adjacent regions the concentration
there will decrease and hence the crystals will dissolve. The process can
produce a series of bands with different crystal sizes and abundances. Both
processes have been combined in a one-dimensional model by Chacron
and L’Heureux (1999). However, observations of the actual formation of
Liesegang bands and layers suggest the post-nucleation model is more
important (John Ross, quoted in Boudreau, 1995). The post-nucleation
process has been proposed for the origin of the famous ‘inch-scale’ layering
of pyroxenes in anorthosite in the Stillwater intrusion (Boudreau, 1995).
Philpotts and Dickson (2002) have observed millimetre-scale layering in
the roof zone of a thick basalt flow. It consists of sheets of ophitic
plagioclase–pyroxene clusters separated by sheets of residual liquid. They
suggest an origin by cycles of nucleation and that such crystal sheets are
separated by gravity along planes of weakness. It is also possible that the
whole structure originates from self-organisation. I have observed that in
some igneous rocks well-developed layering passes laterally into broken layers
and finally clusters. Such behaviour is also predicted by self-organisation
processes (Krug et al., 1996).
203
6.2 Brief review of theory
6.2.3 Grain aggregation and agglutination
Ideally, a grain will sink or rise in a Newtonian fluid at a speed that is related to
the density contrast and size by Stokes’ law. However, in most geological situations the concentration of grains is sufficiently high that grains interact and will
gather into clusters, if there is sufficient time (see the review in Schwindinger,
1999). Such clusters will move as if they were made of a single, larger grain, and
will hence descend faster than each of the original grains. The clusters can form by
vertical and horizontal grain convergence: a large grain will descend faster than
smaller grains and sweep such grains from in front of it. Grains may move
laterally to reduce differences in liquid velocity on either side of the grain – this
is similar to the ‘Bagnold effect’ that produces a concentration of particles in the
centre of a conduit. Grain aggregates may be preserved in volcanic rocks, but it
seems unlikely that such structures can be observed in cumulate plutonic rocks.
In some aggregates euhedral crystals touch along crystal faces (Vance,
1969). One of the best-known examples is that of olivine crystals from the
1959 Kilauea Iki eruption (Figure 6.2; Schwindinger & Anderson, 1989). Most
crystal pairs are united by faces with the same indices, and larger faces are
more commonly in contact than smaller ones. The whole structure resembles a
twinned crystal except that it is not a growth form. The term synneusis
(‘swimming together’) has been used to refer both to nonoriented aggregates
of crystals, and such aligned crystal groups. Schwindinger (1999) considered
that such aligned groups cannot agglutinate during settling, but must be produced during shear flow or turbulence. The driving force of this process is the
reduction of the total surface (interfacial) energy of the mineral grains: contacts
001
101
021
010
110
120
{110}-{110}
{021}-{021}
{010}-{010}
{021}-{010}
Figure 6.2 Synneusis of olivine grains from the 1959 Kilauea Iki eruption
(from Schwindinger 1999).
204
Grain spatial distributions and relations
between similar grains will have a lower surface energy than crystal–magma
contacts. This process has also been proposed for the origin of some garnet
porphyroblasts in metamorphic rocks (Prior et al., 2002).
6.2.4 Magma mixing
If several magmas or crystal mushes are introduced into a conduit or chamber
they will have a tendency to mix. The vigour and duration of the mixing will
determine if the resulting material appears to be homogeneous on a microscopic scale or if the two magmas can still be recognised in blocks and outcrops: the two cases are commonly referred to as magma mixing and magma
mingling. If the original magmas differ in their crystal contents then the
resulting magma may have a non-random distribution of crystals. Of course,
measurements of crystal positions must be done at a higher resolution than
that of the expected structure. In some rocks the results of mixing are clearly
revealed by compositional or isotopic studies. However, if mineral compositions are constant for all grains, then an origin of crystal clusters by incomplete
mixing may not by easy to determine.
6.2.5 Grain packing
Grains in some sedimentary rocks, and crystals in some cumulate igneous
rocks, are packings of hard particles (Rogers et al., 1994). Theoretical packing
of different sized and shaped particles is very complex; hence this subject has
been explored using numerical models of simple shapes, commonly spheres.
These studies have been concerned mainly with the overall packing density,
that is its efficiency, rather than the actual positions of the particles. Jerram
et al. (1996) used one such study to establish a reference texture for nearest
neighbour distances.
6.3 Methodology
The position of a grain has several measurable parameters.
*
*
*
The simplest is its centre, which can be approximated by the centre of mass and
specified by a single triplet of x, y and z values.
The position of the surface of the grain must also be considered: for instance, two
large grains may be distant, but touching. Many more data are needed to specify the
grain surface – many x, y, z triplets.
The nature of a grain’s neighbours: certain mineral pairs may be touching or
spatially associated more frequently than would be expected for a random texture.
6.3 Methodology
205
Most analytical methods assume that the sample is homogeneous. It is particularly important to avoid materials like vesicular lavas, which do not meet
this criterion.
6.3.1 Three-dimensional analytical methods
The centre and surface of a grain can be determined precisely using threedimensional methods, if the surface of the grain can be separated from adjoining
grains by physical or analytical means. In the earliest study of crystal positions a
sample was dismantled using a chisel and the crystal positions measured with a
ruler (Kretz, 1969). However, such a laborious study has not been repeated. The
extent of individual grains can be determined by serial sectioning, if their
concentration is low enough that few crystals are touching, or if grain boundaries are clear (see Section 2.2.1). In transparent materials, such as volcanic
glasses, the extent of crystals can be measured directly with a regular or confocal
microscope (see Section 2.2.2). X-ray tomography can be used to determine the
extent of grains if adjoining grains can be separated (see Section 2.2.3).
These direct methods are the only ones that can be used for grains with
complex shapes. They are also the only ones that can be used to identify
‘chains’ of crystals (e.g. Philpotts et al., 1999). However, it is not always
possible to separate grains using current 3-D methods. Some two-dimensional
methods are much better at separating adjoining crystals and hence may have
to be used (see Section 2.7).
6.3.2 Section and surface analytical methods
In a section the centre of the grain intersection is not the true centre of the
grain, and is therefore referred to as the centroid. For simple shapes the
centroid is the projection of the centre of the grain onto the section. This is
not the case for grains with complex shapes and they cannot be examined
successfully in two dimensions.
Grain outlines can be acquired easily from rock surfaces, slabs and sections,
using a variety of techniques described in Section 2.5. Such images can be
processed manually or automatically to produce classified mineral images (see
Section 2.6). Some methods require that the grain outlines are obtained from
the classified images. The centroid of the grain intersection can be determined
as the centre of mass.
If two grains touch in a section, then clearly they touch in three dimensions.
However, it is not possible to determine from a section if adjacent, but
separate, grains actually touch out of the section plane.
206
Grain spatial distributions and relations
6.3.3 Extraction of spatial distribution parameters
6.3.3.1 Nearest-neighbour distance (‘big’ R)
One of the simplest ways of describing spatial distribution patterns uses the
mean nearest-neighbour distance calculated from the centres of the grains
(Jerram et al., 1996, Jerram et al., 2003). This is compared to the mean
nearest-neighbour distance from a random distribution of points with the
same population density. This is a method that can be applied to any 2-D
textures where a grain or crystal can be defined by a grain centre co-ordinate.
In principle, the method could be extended into 3-D, where the 3-D nearest
neighbour is compared to a random 3-D distribution of points, but as yet very
few 3-D data-sets of rock textures exist and this has not been tested. However,
it is possible to compare the 2-D spatial distribution pattern to existing known
reference textures where their 3-D spatial pattern is known (Jerram et al., 1996,
Jerram et al., 2003). So far this has been limited to isotropic, homogeneous
materials and it is not clear if strongly foliated rocks, or sections with a
significant porosity, can be analysed in the same way.
The positions of grain centres are generally determined when the other
textural parameters are determined. A simple procedure goes through the list
of N crystal centroid positions and finds the distance to the nearest neighbour,
r (see Appendix). The mean nearest-neighbour distance is
X
rA ¼ ð1=NÞ
r
The mean nearest neighbour for a random distribution of points with the same
population density (number of crystals per unit area), NA
pffiffiffiffiffiffiffi
rE ¼ 1=ð2 NA Þ
The ratio of these parameters, ‘big’ R describes the distribution of the grains.
R ¼ rA =rE
In principle, for a distribution of random points R ¼ 1. For clustered points
R < 1 and for ordered points R > 1. However, crystals are not points: each
occupies a certain volume in space and area in a section. Hence, crystals
cannot be coincident like points. To enable the comparison of ‘big’ R values
Jerram et al. (1996) calculated the random sphere distribution line (RSDL,
Figure 6.3). This represents the 2-D ‘big’ R values calculated from 2-D sections
taken through 3-D random distributions of spheres with different porosities
up to the minimum porosity: that for random close packed spheres with a
porosity of 37%. The RSDL is calculated using a combination of computer
207
6.3 Methodology
2.0
Better
sorting
mean distance to nearest-neighbour
distance for random distribution
Mechanical
compaction
Overgrowth
Ra
nd
om
sp
he
re
di
Ordered
st
rib
Clustered
ut
io
n
lin
R=
e
1.0
0
100
abundance of matrix (porosity) (%)
Figure 6.3 Nearest neighbour distance determined from sections through
materials (Jerram et al., 1996). The porosity is equal to 100 abundance in %
of the phase analysed. The random sphere distribution line is indicated on the
graph representing the spatial distributions of randomly packed spheres up to
37% porosity which is the minimum possible for random close packed
spheres.
generated sphere models and a naturally generated model for random close
packing (Jerram et al., 1996). Vectors for common processes have been established, like mechanical compaction, grain overgrowth and better sorting.
This method has considerable advantages as it is simple to apply. However,
the validity of the reference models for more realistic crystal shapes and
variable crystal sizes has not been tested. Also, the method gives only a single
parameter that does not describe cluster size. More elaborate methods will be
described in the following section.
If many or all crystals touch in three dimensions then the resulting framework will have an important effect on the rigidity of the material. Hence there
is much interest in determining the minimum amount of crystals that is necessary to make such a framework. Jerram et al. (2003) used nearest-neighbour
diagrams to determine if crystals in a rock form a touching framework. This
technique has the advantage that it can be applied to sections; however, more
data are needed before the method can be verified fully.
6.3.3.2 Cluster analysis
The goal of cluster analysis is to quantify the size and characteristics of clusters
of grains in a rock. Clustering can reflect packing processes, the distribution of
208
Grain spatial distributions and relations
original nucleation sites or other processes (Jerram & Cheadle, 2000). The
first attempt to quantify clustering sizes and distributions used the nearestneighbour distance technique (see Section 6.3.3.1; Jerram et al., 1996). More
recently two, more complicated, techniques have been developed: complete
linkage hierarchical cluster analysis (CLHCA) and density linkage cluster
analysis (DLCA). These methods are ideally applied to the materials in
which the complete, 3-D position of each grain is known. Jerram and
Cheadle (2000) applied these methods to sections and tested performance
with comparative 2-D sections through known 3-D synthetic textures (SAS
package – see Appendix).
CLHCA was found to give the clearest results. It uses the distance between
the grain centres. These are agglomerated to give the most compact groups.
The result is a dendrogram (tree diagram) that links the grains and groups with
lines. The distance between groups gives the ‘height’ in the tree. However, it is
difficult to get direct quantitative information from a dendrogram – any
distribution will give a dendrogram, including a random distribution.
Another way of presenting the data is a graph of cluster frequency (number
of clusters) against cluster size (number of grains in the cluster). At a cluster
number of one all grains are included; at a cluster size of one grain the number
of clusters equals the number of grains. This is a smooth curve for a random
distribution, but has gentle steps for a clustered distribution. A further refinement can be made by normalising the data to a random distribution; deviations are then easier to see. This can be done by calculating the distance
between clusters and subtracting the expected distance for a random distribution. A graph of this residual distance against cluster frequency gives the
characteristics of the dominant clusters (Figure 6.4).
DLCA uses a different approach: clusters are defined on the basis of the
amount of material within a certain radius. Hence this method considers the
whole volume of the grain, and not just its centre. Clearly, this is a better approach
if grains have a high concentration and many are touching. The technique can be
applied easily to images with pixels or voxels. In many studies we can only observe
grains in section, hence we are faced with the problem of calculating a 3-D
parameter from 2-D data. Jerram and Cheadle (2000) tested this problem with
a synthetic 3-D distribution, as before. They did not achieve good results, but this
may reflect the weakness of the test distribution and not the method.
6.3.3.3 Touching crystals
Analytical methods discussed so far are only concerned with the position of the
grain centre. They are used to reveal processes that acted early on grains.
Conversely, the distance between the edges of the grains will reflect later
209
6.3 Methodology
1.0
Sphene
residual distance
0.5
0
–0.5
–1.0
0
10
20
30
40
50
60
70
cluster frequency (%)
80
90
100
Figure 6.4 Complete linkage cluster analysis of sphene in a granulite sample
(Jerram & Cheadle, 2000). Here the distance between the crystal centres is
compared to that expected for a random distribution (residual distance). The
dotted line is the random variation tolerance limit. Sphene crystals are more
closely spaced than expected in the frequency 5–15%, which corresponds to
cluster size of 1.5 to 5.1 mm.
processes. Obviously this can only be investigated completely in three dimensions. Unfortunately, with most 3-D analytical methods it is difficult to separate
touching crystals and hence there has been no work on this subject. If the surface
of contact between two grains is not parallel to the section plane, as is usually the
case, then some, but not all, touching grains will be identified in sections.
Clusters of touching grains are frequently observed in volcanic rocks. In
some cases the touching grains are sub-parallel (synneusis). Although it is
possible to quantify this effect by looking at grain edge distance against
differences in orientation, this has yet to be done.
Some rocks contain a touching network of plagioclase crystals that are
visible in sections as a collection of crystal chains (Philpotts et al., 1998). In
three dimensions these chains must be projections of foam-like surfaces of
crystals, but it is easier to continue the use of the term chain. The crystal chains
are revealed most clearly when a block of basalt is melted: the block retains its
shape even when 70% is liquid. Philpotts et al. (1998) identified the chains
manually from reconstructed serial sections and X-ray tomography. There are
many aspects of the chains that can be quantified, but so far the most useful
has been the number density of the chains (Philpotts et al., 1999). This was
determined by counting the number of chains intersected along various random straight lines in sections. The number density can be transformed easily
into the mean chain spacing, also called the network mesh size. The orientation
of such chains is discussed in Section 5.8.3.1.
210
Grain spatial distributions and relations
6.3.3.4 Wavelet transform and Fourier analysis
The wavelet transform is a powerful method to identify ordered structures in
rocks (Gaillot et al., 1997, Gaillot et al., 1999). It can be used to identify the
orientation and spacing of ordered structures and has been discussed in the
section on grain orientation (see Section 5.4.3.3). Fourier analysis of grain
sections has been used to look at grain distribution in sandstones (Prince et al.,
1995). Both these methods examine an image of the distribution of a mineral or
grains, not individual grains.
6.3.4 Extraction of mineral association parameters
6.3.4.1 Mineral point correlations
Morishita and Obata (1995) proposed a rather different way of looking at the
spatial distribution of grains and the relationship between grains of different
minerals. It is based on the frequency of occurrence of pairs of points in a
section with the same or different minerals versus distance. It uses a classified
mineral map, but it is not necessary to define individual mineral grains. The
method can only be applied where there is more than one phase in the rock.
The method could also be applied in three dimensions.
We will consider first a homogeneous, bimineralic rock, for simplicity. If
two points are randomly selected then the unconditional probability that both
are mineral 1 is V12, where V1 is the volumetric abundance of phase 1.
Similarly, the probability that both are mineral 2 is V22. The probability that
we have chosen a mixed pair is 2V1V2. We next consider the conditional
probability of occurrence of mineral pairs at a specified distance D. These
depend on the grain size and spatial distribution of the grains and are denoted
P11(D) for a pair of points of mineral 1, P12(D) for a pair of points of mineral 1
and mineral 2, etc. For D ¼ 0 the points are coincident and the probability is
equal to the volumetric abundance: P11(D) ¼ V1. Similarly the probably that
the points are different is zero: P12(D) ¼ 0. Where D is very large the probabilities become equal to the unconditional probability, that is P11(D) ¼ V12 and
P12(D) ¼ 2V1V2. The difference between the unconditional and conditional
probabilities expresses the grain size, degree of order and degree of association
of the mineral pairs. It is useful to define textural parameters that express these
differences and extend the treatment to more than two minerals. For each
mineral i a parameter si (D) is defined for a distance D:
si ðDÞ ¼
Pii ðDÞ V2i
Vi V2i
6.3 Methodology
211
For each mineral pair i, j a parameter tij (D) is defined for a distance D:
tij ðDÞ ¼
Pij ðDÞ
2Vi Vj
At D ¼ 0 all s ¼ 1 and t ¼ 0; at D ¼ 1 all s ¼ 0 and t ¼ 1.
Morishita and Obata (1995) examined a granite and a hornfels using these
methods. They found that s values tend to decrease exponentially with distance and the slope is a measure of mean crystal size. If the rock was ordered
strongly then the s values should peak at that distance, but this behaviour has
not been recorded. They also found that t values tend to increase monotonically with distance. Some mineral pairs rise to a peak and then go down,
expressing a spatial association of those two minerals and giving a typical
mineral spacing.
Morishita (1998) developed these measurements further by comparing
s and t values of an ‘ideal rock texture’. In this all crystals are randomly
( uniformly) distributed in the rock. He showed that if a number of assumptions are made then the crystal size distribution can be calculated from the
variation in the probability that a point lies within a crystal against the distance
from the crystal centre. However, it is not clear if this technique has any practical
use, as it is not obvious how close natural rocks lie to this ideal texture.
This technique could certainly give interesting information on mineral
associations, data which are not given by other methods. However, the lack
of a readily available computer program to calculate the textural parameters
has hampered the application of the method.
6.3.4.2 Box-counting method
Mineral associations can also be detected using a box-counting method
(Saltzer et al., 2001). This starts with a classified mineral image of a section
of a rock. The image is first divided into four square areas each with an area
of 22n pixels where n is the largest integer possible allowed by the image size.
The value of n is decreased to zero, where each box contains one pixel. For
each value of n the numbers of boxes dominated by two minerals are
tabulated for each mineral pair. This is compared with a synthetically
derived random distribution. A significant difference in the distributions
indicates that the mineral pair is associated. This method does not quantify
the correlation, but just indicates significant associations. The method
does not seem to have any advantages over the method of Morishita and
Obata (1995) and suffers from the same lack of an easy tool for doing the
measurements.
212
Grain spatial distributions and relations
6.4 Typical applications
One of the problems with this subject is that there is a large amount of
theoretical and modelling work that produces patterns that resemble rocks
(e.g. Ortoleva, 1994, Jamtveit & Meakin, 1999), but few actual measurements
of the rocks themselves. Therefore, commonly, it has not been possible to
verify if the proposed pattern-making processes actually occurred. In addition,
many of the techniques presented in papers have not been taken up by other
workers; hence the only example is the original paper. In some cases this is
clearly a result of a lack of an easy-to-use computer program. A clear exception
to this trend is the ‘R’ parameter of Jerram et al. (1996).
6.4.1 Metamorphic rocks
Kretz (1966a) examined both the size distribution and position of garnet and
biotite in a schist, and phlogopite, pyroxene and titanite in a marble. He
measured the 3-D position of the crystals by disintegrating the samples with
a small chisel. He found that the spatial distribution of garnet crystals was
random and their size did not correlate with the distance to their nearest
neighbour. Three years later he examined a section of pyroxene–scapolite–
titanite granulite in more detail (Kretz, 1969). He divided up the section into
sub-areas and measured the number density of crystals in each sector. By
comparing the data to the expected abundance he was able to use a chi-squared
test to show that the crystals were distributed randomly. He also showed that
titanite was not associated preferentially with any of the major phases. He
concluded that nucleation of all phases was random. The same section was
re-analysed by Jerram and Cheadle (2000) using their cluster-analysis method.
They found that the larger crystals of scapolite and pyroxene act as ‘domains
of order’ – that is, there are fewer small nearest-neighbour distances than
expected (see Figure 6.4). This method also showed that titanite is clustered.
6.4.2 Igneous rocks
Jerram et al. (2003) have reviewed data on clustering using a graph of their R
parameter against per cent porosity (Figure 6.5). They find that crystals from a
rhyolite laccolith are distributed randomly (see below; Mock et al., 2003), but
that all other examples are clustered. They established a field on this graph for
touching frameworks by the inspection of samples. A good number of clustered examples with known touching frameworks have been used to define this
region, but few non touching frameworks have been analysed, so perhaps
213
6.4 Typical applications
Belingwe komatiite (Joe’s Flow)
1.9
?
Belingwe basal chill (Joe’s Flow)
Alexo
Kambalda
ORDERED
?
1.7
Komatiites
Vammala
Campbell et al. (1978) experiment
Alexo
RS
1.5
R
Vammala
Holyoke Colonnade
DL
Holyoke Entablature
Mock et al. (2003) data
Campbell
1.3
RSDL = Random sphere
distribution line
Kambalda
1.1
Belingwe komatiite
Non-touching crystals
(random to ordered
away from the RSDL)
Entablature
Non-touching crystals
(random to clustered
away from the RSDL)
Belingwe
chill
0.9
CLUSTERED
Holyoke Colonnade
Touching framework
0.7
0
20
40
60
porosity (φ) (% melt)
80
100
Figure 6.5 Porosity (%) or percentage melt against R value (from Jerram
et al., 2003).
more data are needed to establish this field more securely. Ikeda et al. (2002)
examined two samples of different granites using the method of Jerram et al.
(1996) and concluded that all minerals were clustered, except for biotite in one
sample.
Boorman et al. (2004) examined the positions of orthopyroxene and plagioclase crystals in mafic rocks from the lower part of the Bushveld complex using
the nearest-neighbour approach of Jerram et al. (1996). The R against porosity
diagram was developed for spheres, but despite this limitation Boorman et al.
(2004) considered that it could be applied to tabular crystals, such as those
measured here (Figure 6.6). The vector defined by their samples could be the
result of sorting, but does not correlate with the aspect ratio of the grains or the
slope of the CSD, as would be expected. Instead, Boorman et al. (2004) suggest
that this vector represents deformational compaction, similar to that proposed
by Meurer and Boudreau (1998).
The spatial distribution of plagioclase, orthoclase and quartz in a rhyolite
laccolith were examined by Mock et al. (2003). The more tabular feldspars
seemed to show greater variation than the more equant quartz (Figure 6.5).
There was a trend to clustering towards the top of the laccolith, and internal
pulses of magma were distinguished.
214
Grain spatial distributions and relations
Mechanical
compaction
Overgrowth
2
Group 1 opx
Better
sorting
Group 2 opx
Group 3 opx
R-value
Deformational
compaction
Group 4 opx
Group 4 plag
Aging
RSD
L
Ordered
1
Clustered
0
50
% porosity
100
Figure 6.6 Spatial distribution patterns of orthopyroxene (squares) and
plagioclase (circles) in the lower part of the Bushveld complex (Boorman
et al., 2004). Shaded region is for touching frameworks.
The more complex cluster analysis method of Jerram and Cheadle (2000)
was applied by them to two samples. They showed that in a komatiite sample
olivine crystals are clustered with diameters of 0.3–2.5 mm. In the scapolite–
pyroxene–titanite rock examined by Kretz (1969) they found that titanite was
clustered on a scale of 1.8–5.1 mm.
Chains of plagioclase crystals appear to be an important petrological structure in the thick Holyoke basalt lava flow and other slowly cooled bodies of
similar composition (Philpotts et al., 1999). The chains are a few crystals wide
and consist of crystals up to 0.5 mm long attached in random orientations.
They have been traced by hand from thin sections (Figure 6.7). They have been
quantified by counting the number of intercepts per unit length and appear to
be most numerous at the base of the flow and diminish upwards. The chains
appear to originate in the upper part of the flow as groups of plagioclase
crystals embedded in and surrounding pyroxene oikocrysts (Philpotts &
Dickson, 2000). These crystal ensembles descend by convection to the lower
part of the flow where the pyroxene exsolves and recrystallises into masses of
granular pyroxene surrounded by the plagioclase chains. An investigation of
the chains with a view to determining anisotropy produced by compaction is
discussed in Section 5.8.3.1 (Grey et al., 2003).
Igneous layering is commonly observed and has been the subject of numerous theoretical and qualitative studies (Parsons, 1987, Naslund & McBirney,
1996). However, so far there has been little quantitative work on crystal
215
6.4 Typical applications
Centre of 174 m-thick flow
58.7 m
compacted
Vertical sections
74.2 m
40.9 m
Increasing mesh size
dilated
2 cm
19.9 m
No chains present in first few metres
Flow base
Figure 6.7 Plagioclase crystal chains in the Holyoke basalt flow (Philpotts
et al., 1999). Chains were traced by hand from thin sections.
positions in layered plutonic rocks. A good example is the inch-scale layering
in the Stillwater intrusion. Boudreau (1987, 1995) has done a fine job of
modelling the origin of the layering by self-organisation, but there are no
quantitative studies of the grain positions and sizes in the rock.
There have been few studies of mineral associations: all are by the authors
that originated the method and all just indicate significant associations without giving a value to this parameter. The spatial association of garnet and
clinopyroxene in a sample of mantle peridotites was considered to be produced
by mineral breakdown: orthopyroxene reacted to form garnet and clinopyroxene (Saltzer et al., 2001). This was used to reconstruct a pressure–temperature
history for these samples. Morishita and Obata (1995) examined two samples
of hornfels and granite and found that biotite and quartz were spatially associated in the granite.
7
Textures of fluid-filled pores
7.1 Introduction
Many igneous and metamorphic rocks contain a fluid component or evidence
of the former existence of such a fluid.1 Fresh lavas and pumice blocks
commonly contain vesicles, formed by the exsolution of gas from the silicate
fluid. Metamorphic rocks may preserve evidence of partial melt pockets. The
term ‘pore’ is used here to indicate space filled, or formerly filled, with fluid
surrounded by a more viscous fluid (e.g. silicate melt, glass) or crystals. Pores
may be considered as the fluid equivalent of grains or crystals in solid phases.
Fluid inclusions in crystals are excluded here as their textures have yet to be
studied quantitatively.
Pores do not always stay filled with their original fluid during lithification.
For instance, vesicles may be filled with other minerals to become amygdules.
In partially molten silicate rocks the fluid may quench to a glass, or a mixture
of minerals, whose textures pseudomorph the original pores. Hence the porosity of a rock, as discussed here, is the original volume proportion of fluid at
the time of interest. For example, the porosity of a lava sample is the volumetric proportion of vesicles at the moment of solidification; the porosity of a
partially molten rock is the volumetric proportion of fluid when the texture is
preserved and such a melt may have gas bubbles. Clearly, it is important to
specify what feature is being examined. The porosity is characterised at a first
approximation by the total volumetric proportion of fluid, but like grain size,
it is also possible to quantify the size distribution of pores.
Surfaces are important in many geological processes, as sites of chemical or
mechanical processes. The pores define an internal surface area in a rock. The
outside of grains defines the external surface area, which may be significant for
loose materials. Clearly some materials, such as pumiceous pyroclastic deposits will have significant internal and external contributions to the total surface
216
7.2 Brief review of theory
217
area. In natural materials surfaces are complex, hence the actual value of the
total surface area depends partly on the scale of the analysis and the measurement method (Brantley et al., 1999, Brantley & Mellott, 2000).
Another important associated property is the permeability of the rock. In
many ways it is much more important than the porosity, as this parameter
measures how readily a fluid can flow through a rock. The total permeability
can be measured, in addition to the size distribution of pore channels and
‘throats’ that control fluid flow. Rocks with a high porosity tend to be highly
permeable, but the porosity–permeability relationships diverge as the porosity
goes down.
7.2 Brief review of theory
7.2.1 Permeability
The permeability of a rock is used qualitatively to describe the ability to
conduct fluid: a more permeable rock conducts fluid more easily that a less
permeable rock. Hence the permeability expresses the connectedness of the
pores. The original development of the subject considered the flow of water in
a rock and is expressed as Darcy’s law:
DP=l ¼ m=k
where DP is the pressure drop over a length l, is the specific discharge or filter
velocity, m is the dynamic viscosity of the fluid and k is the permeability. The
permeability is sometimes expressed in Darcy units, which are equal to about
108 cm2. The permeability of a rock is not necessarily the same in all
directions.
Darcy’s law is only applicable in low Reynolds number (ratio of inertial
forces to viscous forces) flow where energy loss is due to viscous effects. At
higher flow rates the fluid inertia becomes important and another term must be
added to Darcy’s law (see review in Rust & Cashman, 2004):
DP=l ¼ m=k1 þ 2 r=k2
where k1 is the viscous (Darcian) permeability, r is the fluid density and k2 is
the inertial (non-Darcian) permeability.
7.2.2 Magmas and partially molten rocks
Movement of fluid with respect to solids is a fundamental process in the
solidification of igneous rocks and the melting of metamorphic rocks. It may
218
Textures of fluid-filled pores
occur by compaction (McKenzie, 1984), porous-media convection or in
response to shear. It is especially important in the production of monomineralic rocks, such as anorthosite or dunite. The final porosity is commonly
expressed as the ‘trapped liquid’ component and can be determined by chemical or textural methods. The key property is clearly the variation in permeability during solidification or melting: if the permeability is finite for all
values of porosity, then all the silicate fluid can leave the rock. However, it is
commonly proposed that there is a percolation threshold, below which the
permeability falls to zero. The actual value of such a threshold may be partly
controlled by the interfacial angles of the fluid with the solid phases, which in
turn express differences in interfacial energy (see Section 4.2.3).
Cheadle et al. (2004) have reviewed and investigated the relationship
between textural equilibrium, dihedral angles and permeability using a
geometrically simple numerical model. The simplest models are for settled or
random rigid shapes. These are termed non equilibrated as their shape is
unchanged after accumulation. For these materials the percolation threshold
varied with crystal shape – it was about 3% for spheres and 10% for
cubes (Figure 7.1). This means that there will always be a minimum of 3 to
10% melt trapped in the cumulate. For texturally equilibrated samples, with
dihedral angles less than 408, the percolation threshold was close to zero.
That is, all interstitial fluid (silicate liquid) can escape and monomineralic
rocks can form.
0
el
em
ub
cub
em
od
el
od
Rand
om
c
10–14
°)
Settle
d
E
10–12
40
(θ <
ted
a
r
s
ilib
re
qu
he
sp
10–10
Se
ttle
d
permeability, k, (mm–2)
10–8
10
20
30
porosity (%)
Figure 7.1 Melt permeability against porosity for a crystal–liquid system.
For the settled and random models the porosity is the space between crystals
represented as simple geometrical forms. The crystals are allowed to interact
in the equilibrated model, giving dihedral angles, , of less than 408. From
Cheadle et al. (2004).
7.2 Brief review of theory
219
7.2.3 Bubbles in magmas
7.2.3.1 Nucleation and growth of bubbles
Bubbles nucleate in a similar way to crystals (Figures 7.2a, 7.2b). Nucleation
occurs when the gas becomes supersaturated in the magma. Classical nucleation theory (e.g. Gardner & Denis, 2004) predicts that the activation energy,
DF, for formation of a bubble nucleus in a homogeneous fluid is
DF ¼
16ps3
3DP2
where s is the surface tension of the bubble, and DP is the supersaturation
pressure.
This equation must be modified for heterogeneous nucleation (on a crystal
surface): s becomes the balance of surface tensions for crystal, melt and bubble
and the activation energy is modified by a factor f which is related to the
wetting angle (solid–liquid dihedral angle; see Section 4.2.3) of the bubble on
the surface (; Figure 7.3): f ¼ 1 for completely wetted surfaces (identical to
homogeneous nucleation) and f ¼ 0 for a wetting angle of 1808 (Gardner &
Denis, 2004). Application of this simple theory predicts that homogeneous
nucleation requires a high degree of supersaturation and hence may be rare.
However, heterogeneous nucleation requires a much lower degree of supersaturation, and hence will dominate if crystals are present. For example, iron
oxide crystals appear to be better nucleation sites as bubbles do not wet the
surface, whereas plagioclase is wetted by bubbles and hence requires a very
high supersaturation pressure (Figure 7.3; Hurwitz & Navon, 1994, Gardner &
Denis, 2004).
After nucleation bubbles grow by diffusion of vapour through the liquid
(Figure 7.2c). The speed of this process will depend on the viscosity of the
liquid and the molecular weight of the vapour: diffusion is fastest for light
molecules in a low-viscosity fluid. There is a feedback between bubble growth
and nucleation of new bubbles. Growth of bubbles by diffusion reduces the
vapour content of the melt, which will in turn affect the surface tension. A
lower water content in the melt will require a higher supersaturation pressure;
hence nucleation of new bubbles will be more difficult. The overall result of
this feedback is that nucleation may be instantaneous, rather than continuous.
Once bubbles have formed they will increase in size when the magma moves
to a lower pressure region (Figure 7.2d). Expansion will be slowed down by
viscous effects and relatively large changes in depth are probably needed to
make significant changes in bubble size.
220
Textures of fluid-filled pores
(a) Homogeneous
nucleation
(b) Heterogeneous
nucleation
(c) Growth by
diffusion
(d) Expansion owing to
pressure changes
(e) Coalesence
(f) Coarsening
(Ostwald ripening)
Figure 7.2 Bubble formation, growth and shrinking. (a) Homogeneous
nucleation occurs within a fluid. (b) Heterogeneous nucleation occurs on a
surface. If the angle between the bubble and the surface is small the surface is
said to be wetted and nucleation requires a higher degree of supersaturation.
(c) Growth of pores can occur by diffusion of fluid into the pore. (d) If the
overall pressure is reduced bubbles will expand provided the matrix has a low
enough viscosity. (e) Bubbles may also grow by coalescence. (f) Small bubbles
have a higher internal pressure than larger bubbles. As the material
approaches equilibrium vapour will be transferred via the matrix, leading to
a coarsening of the bubbles.
The kinetically controlled processes of bubble nucleation and growth can
continue as long as there is sufficient supersaturation. As the driving force for
these processes diminishes, bubbles will try to approach equilibrium by size
adjustments. Bubble coalescence occurs when two bubbles touch and become
one: it has no parallel in crystal growth (Figure 7.2e). Rates of coalescence
depend on the timescale of bubble approach, film thinning, rupture and
7.2 Brief review of theory
θ = 30°
(a) Surface wetted – high ΔP
(e.g. plagioclase)
221
θ = 120°
(b) Surface not wetted – low ΔP
(e.g. magnetite)
Figure 7.3 Nucleation bubbles on surfaces. (a) Wetted surfaces have low
dihedral angles ( ¼ 308) and require large supersaturation for nucleation.
(b) Unwetted surfaces ( ¼ 1208) require lower degrees of supersaturation.
relaxation. Experimental results suggest that coalescence is most important
when the bubbles are expanding in response to a reduction in the confining
pressure (Figure 7.2d; Larsen et al., 2004). Gaonac’h et al. (1996a) have proposed that cascading bubble coalescence dominates other equilibrium processes.
During the final stages of development cascading bubble coalescence can lead to
catastrophic fragmentation of the magma (Gaonac’h et al., 2003).
Coarsening (Ostwald ripening) of bubbles resembles the same process in
crystals (Figure 7.2f; see Section 3.2.4). A larger bubble has a lower internal
pressure than a smaller bubble. Hence, gas will move from the smaller bubble
to the larger by diffusion through the silicate liquid. Clearly, the importance of
this process depends on the rate of diffusion of gas through the silicate liquid,
which in turn is controlled by the distance, temperature, liquid viscosity and
gas composition. However, unlike coalescence, it will occur even if the bubbles
do not intersect (Larsen & Gardner, 2000).
7.2.3.2 Bubble size distributions and shapes
Bubble size distributions (BSDs) can follow many different models, such as
unimodal, polymodal, exponential and power law (Blower et al., 2002,
Gaonac’h et al., 2003). Unimodal BSDs are produced by instantaneous bubble
nucleation, most likely following a heterogeneous model (Blower et al., 2002).
Polymodal BSDs probably form by several instantaneous nucleation events.
Both exponential and power law BSDs are commonly observed. The similarity of exponential BSDs to some CSDs (Marsh, 1988b) has suggested to
some authors that they can form in the same way, by continuous nucleation
and growth under steady-state conditions (e.g. Mangan & Cashman, 1996).
Power law BSDs may form by cascading coalescence of bubbles (Gaonac’h
et al., 2003). However, other models are possible: Blower et al. (2002) have
proposed a model of non-equilibrium continuous nucleation and growth of
bubbles that evolves from a unimodal distribution via an exponential model to
a power law distribution.
222
Textures of fluid-filled pores
The shape of bubbles in magma can be changed by deformation (Rust et al.,
2003). These shapes will be preserved if the relaxation times of the magma are
long, as is the case for a viscous magma such as rhyolite. However, nonspherical bubbles are not usually preserved in mafic magmas, except near the
chilled surface. The shape of bubbles has been used to quantify the type of
shear, simple or pure, and the shear rate (Rust et al., 2003).
7.3 Methodology
The overall porosity and pore size distribution of a material can be measured
using 3-D and 2-D methods, to be discussed below. Pores are commonly
assumed to approximate to spheres, hence their size is considered to be the
diameter of a sphere of equivalent volume (Figure 7.4). If the pores are
connected then the size of the pore is calculated assuming that the pore is
closed at the pore throat. For some strongly anisotropic materials pores may
depart from such a model and then size definitions similar to those used for
grains may be used (see Section 3.2.1.1). For some applications the actual
volume of the pore is used instead of the diameter.
The permeability of a material depends on the connections between the
pores (Figure 7.4). Connections may be increased by fracturing of the material
Indeterminate
pores
Isolated
pores
Connected
pores
Equiv pore
diameter
Throat width
Figure 7.4 Schematic 2-D section of a porous material. Pores were or are
filled with fluids, such as oil, gas, water or silicate melts. Pores may be divided
into those that are isolated (solid edges) and those connected by throats
(dashed edges). This is a section so isolated pores may be connected in three
dimensions outside the plane of the section. It is not possible to classify pores
that intersect the edge of the sample (dotted edges).
7.3 Methodology
223
during preparation. If the sample is small then permeability may be dominated
by edge effects: all pores that intersect the surface of the sample are connected
to it. However, their connections in the original rock are unknown. Hence, a
researcher should be aware of connections between the size of the sample and
the resolution of the method.
7.3.1 Pore surface area measurements
7.3.1.1 Analytical methods
Nitrogen adsorption is a physical method for measuring the total surface area
of a material permeable to gas (Gregg & Sing, 1982, Whitham & Sparks, 1986).
It cannot measure the area of isolated pores. The sample is first dried in a
vacuum. Then a stream of nitrogen is passed over the sample and allowed to
come to equilibrium. The nitrogen is adsorbed as a single layer of molecules on
the surfaces. Pure helium is then passed over the sample and the nitrogen
releases from the surface. The amount of nitrogen in the helium is measured.
The experiment is repeated for different partial pressures of nitrogen. From
these data the surface area per unit volume can be determined. If a shape for
the pores is assumed, such as a cylinder, then the mean pore diameter can be
calculated.
Mercury intrusion porosimetry involves the adsorption of mercury by a
permeable rock under increasing pressure (Whitham & Sparks, 1986). The
adsorption is opposed by the surface tension of the mercury. This in turn
depends on the wetting angle and the diameter of the pores. The geometry of
the pores must be assumed – a cylinder is the most common shape used. The
measurements are done in two apparatuses: the sample is first put in a vacuum
and mercury adsorbed, typically at a pressure of 0.078 bar, which corresponds
to a pore diameter of 235 mm. The pressure is increased to atmospheric,
equivalent to a diameter of 15 mm. Finally the sample is transferred to a
pressure vessel where the pressure is gradually increased: a pressure of
4280 bar corresponds to a pore diameter of 0.0032 mm. This method is useful
for the smaller pores, but again only counts pores connected to the outside of
the sample. However, it can be expected that thin pore walls will fracture under
these high pressures. Total porosity can also be determined in a similar way by
adsorption of water under pressure (Whitham & Sparks, 1986).
7.3.1.2 Image analysis method
The surface area per unit volume is one parameter that can be determined
exactly from intersection data (see Section 2.7.2). Although this method is
224
Textures of fluid-filled pores
independent of pore shape, connectedness or orientation, it is best to use a
number of orthogonal sections for markedly anisotropic materials. The sections
are imaged using optical or electronic methods (see Section 2.2). Test lines are
drawn on the images and the number of intersections of the line and the pores
counted. The surface area per unit volume is equal to twice the number of
intersections divided by the length of the test line. This method counts all
pores, including the isolated pores missed by the nitrogen adsorption method.
7.3.2 Three-dimensional analytical methods
7.3.2.1 Total porosity
Total porosity can be calculated easily from the density of the rock, that of the
void-free rock and the void-filling material. The density of the rock is most
easily determined by cutting an orthogonal block and measuring and weighing
it. Archimedes’ principle can also be used provided that the measuring fluid,
usually water, does not get into the pores (Houghton & Wilson, 1989). If the
pores are less than 1 mm then the block can be sprayed with waterproofing
aerosol as used on camping gear. For larger pores the block can be wrapped in
wax sheet, whose mass must be considered. Rocks which are less dense than
water can be ballasted with a known weight. For rocks with a known mineralogical composition the void-free density can be calculated by summing the
proportion of the minerals weighted for their density. For igneous rocks the
density can be estimated from the chemical composition (Ghiorso & Sack,
1995). This parameter is calculated by a number of geochemical programs (see
Appendix).
The true rock density can also be measured by helium pycnometry: the
volume of helium displaced by the sample increases the pressure in a chamber
and the density can be calculated using the ideal gas law (Klug & Cashman,
1994). Helium will only enter into all pores connected to the surface hence the
total porosity may be less than the density method described above. The
measurements are done in a dedicated instrument.
7.3.2.2 Three-dimensional imaging techniques
The distribution of pores can be determined by serial sectioning (see
Section 2.2.1). In transparent materials, such as volcanic glasses, the size of
bubbles can be measured directly with a regular or confocal microscope (see
Section 2.2.2).
X-ray tomography methods similar to those used for measuring the distribution of crystals can be used for pores also (see Section 2.2.3). Synchrotron
7.3 Methodology
225
X-ray microtomography is very similar, but uses a more intense source with a
finer resolution (Song et al., 2001, Shin et al., 2005). Finally, magnetic resonance imaging has been used to visualise the texture of carbonate sediments
(Gingras et al., 2002). This method is particularly sensitive to water distribution. However, the resolution of the method is poor and there are few advantages over other methods.
7.3.3 Two-dimensional analytical methods
7.3.3.1 Measurement of pore outlines
The most popular method for determining porosity and pore size distributions
is image analysis of sections or surfaces. The open spaces in rocks may be made
more visible in thin section if the rock is first impregnated under vacuum with
coloured epoxy resin. This process clarifies the distinction between isotropic
materials like volcanic glass and the open pores. In slabs it is better to use a
fluorescent dye. With this method pores down to 1 mm can be imaged. Some
workers have used low-melting-point metals for impregnation, such as Woods
metal. Unfortunately this alloy is rich in cadmium and hence toxic.
Sections can be imaged in transmitted or reflected light (see Section 2.5).
They can also be examined using a scanning electron microscope. Images can
be processed and segmented using the same methods as for minerals (see
Section 2.3).
7.3.3.2 Conversion of two-dimensional data to 3-D pore parameters
Sectional data must be converted to 3-D data using stereologically correct
methods, such as those developed for grain and crystal intersections (see
Section 3.3.3). The spherical or sub-spherical shape of many pores, especially
in volcanic rocks, makes correction for the cut-section effect less significant
than for more complex shapes. However, the intersection-probability effect
remains just as strong.
Small pores (<30 mm) may be seen in projection when examined in thin
section. In this case the equations for conversion of data are different from
those used for intersection (see Section 3.3.4). For spherical pores the data do
not need any correction at all. Gardner et al. (1999) used a variant of the
projection method: they counted the number and size of bubbles within a small
volume and then used the total bubble volume to scale the measured bubble
size distribution to the values for unit volume.
Thick sections with smaller pores viewed in projection and larger pores
viewed in intersection are difficult to treat. Ideally the two populations must
226
Textures of fluid-filled pores
be separated, calculated separately using the appropriate methods and then
recombined.
Blower et al. (2002) have proposed a parametric model for conversion of
sectional data. They note that if the pores are spherical and their sizes follow a
power law distribution, then the distribution of sectional diameters will also
have a power law distribution. Hence, the parameters of the 3-D power law can
be determined from the 2-D power law. Parametric solutions are always
dangerous to use as pore and grain sizes rarely conform to predetermined
distribution laws. In fact, we are generally searching for the distribution and
hence it is unwise to assume it beforehand (see Section 3.3.4).
Some early studies used the ‘Wager’ equation to convert intersection size
data to pore size distributions (e.g. Mangan et al., 1993). As was discussed in
Section 3.3.4.6 this equation is not appropriate, and does not give correct
results, but published results can commonly be recalculated using more accurate methods (Higgins, 2000).
Toramaru (1990) proposed the use of the ‘Spektor’ method for determining
the pore size distribution. This method gives correct results, but necessitates
measurement of a very large number of pores to give precise results. It is
usually better to measure pore intersection size directly and make stereological
corrections.
Pore shapes and orientations can be determined using the same methods as
for solid grains (see Sections 4.2 and 5.2).
7.3.4 Measurement of permeability
Permeability is generally measured in a dedicated instrument. For example, for
a volcanic rock the sample is a cylinder of rock, typically 25 mm in diameter
and 23 mm long (Rust & Cashman, 2004). The edge of the core is sealed with a
rubber sleeve, leaving the circular ends open. Gas is forced through the sample.
The pressure difference between the two ends is increased in many small steps
from zero to 1.4 bar and the gas flow rate measured. From this the permeability parameters for the viscous and inertial components of the permeability
can be determined.
7.4 Parameter values and display
7.4.1 Display of pore size distributions
Pore size distributions (PSD) can be expressed and displayed in a number
of ways. Some of these are similar to those used for grain size distributions
7.4 Parameter values and display
227
(see Section 3.3.7). As for grain size distribution, it is always important to
recognise the resolution limits of any analytical method. These will clearly
affect the interpretation of the pore size distribution and the values of parameters that describe the pores.
A diagram similar to that used for crystal size distributions can be used (see
Section 3.3.7.2; Marsh, 1988b). In this graph the natural logarithm of the
population density is plotted against size, generally expressed as the equivalent
radius or diameter. As before population density must be distinguished clearly
from volume number density (see Section 3.3.7.2). This diagram is useful if the
pore population is generated by nucleation and growth in a similar way to
crystals. However, in many situations nucleation is instantaneous and hence
the Marsh model is not applicable.
Many authors have used simple frequency histograms to display pore size
distributions. The horizontal axis is either size as equivalent radius or volume
expressed linearly or in log 2 units (Toramaru, 1990). However, as was mentioned in Section 3.3.7, this is not a very good diagram if the data are sparse, as
they tend to be for the larger size classes. In this situation many classes will only
contain one or two pores or may be empty (e.g. Larsen et al., 2004). The eye
tends to glide over the empty classes, creating the impression that the number
of larger pores is greater than it is. This problem cannot be resolved by
increasing the width of classes as then the height of the histogram bar will
change. A better, and equally simple, diagram uses frequency density. The
frequency in each size class is divided by the width of the class. Larger classes
can then be wider, eliminating empty or sparse size classes that are just
artefacts of measurement.
Other distribution models can be verified by suitable diagrams in the same
way as for crystal size distributions (see Section 3.3.7). For example, Klug et al.
(2002) used a cumulative frequency diagram to verify if BSDs were lognormal
and a log–log diagram to determine if a power law (fractal) distribution was a
better approximation.
The connectedness of pores can also be quantified. It can be expressed in
terms of the number density (number per unit volume) of interconnected pores
against pore volume (Song et al., 2001). Pores are connected by pore throats
and the dimensions of such throats can also be expressed by a mean value and
size distribution.
7.4.2 Overall pore size parameters
The moments of the pore size distribution, M0, M1 etc. can be calculated easily
and are simply related to the overall properties of the pores. These equations
228
Textures of fluid-filled pores
are slightly different from those in Section 3.3.4 as the pores are assumed to be
spheres and the size variable is the radius of the pores.
M0 = total number of pores per unit volume, NV.
M1 = total radius of pores ¼ NV the mean radius of the pores, Rm.
M2 = total projected area of pores ¼ 1/4p surface area per unit volume, SV.
However, SV can be obtained precisely by other methods that do not need to
assume a pore shape (see Section 2.7.2).
M3 = total volume of pores ¼ 3/4p porosity. Again, the total porosity can be
obtained precisely by other methods that do not need to assume pore shape
(see Section 2.7.2).
7.4.3 Pore shapes
The shape of pores can be quantified in the same way as for crystals and grains
(see Section 4.3.3). However, simple aspect ratio parameters are used most
commonly. Small vesicles commonly have the form of an ellipsoid and hence
their shape can be specified by their aspect ratio (e.g. Rust et al., 2003). Pore
orientation can again be quantified in the same way as for grains (see Section 5.5).
A commonly used measure is the orientation of the long axis of the pore with
respect to a structurally important direction, such as flow for lavas. The overall
texture of pores can also be described in terms of fractals (Meng, 1996).
7.5 Typical applications
Pore size distributions (PSDs) have been measured in a wide variety of rocks.
Some of the early studies that determined PSDs from measurements made on
sections used incorrect conversion methods; hence the data must be reconverted before it can be compared with measurements made in three dimensions
or using more accurate stereological techniques. The equivalent problem for
crystals is discussed in Section 3.3.4.6.
Partially molten rock contains silicate-liquid-filled pores and can be studied
in the same way as other porous materials. In some cases the liquid is preserved
as a glass, but in many plutonic rocks its presence must be inferred from the
textures of minerals considered to pseudomorph the original melt-filled pores.
So far pore structure has been explored by measurement of dihedral angles of
solid–solid–liquid intersections and these results are discussed in Chapter 5.
7.5.1 Volcanic rocks
Quantitative study of vesicles is a very active research field of volcanology
because nucleation and growth of bubbles can trigger and drive volcanic
229
7.5 Typical applications
eruptions (e.g. Sparks, 1978). Vesicle size distributions (VSDs) are studied
because they may give information on degassing and magma volatile contents
(Toramaru, 1989, Gaonac’h et al., 1996b, Larsen & Gardner, 2000) and may
be used to determine eruption conditions (Blower et al., 2001).
Explosive silicic volcanic eruptions are not only important for their geological effects, but also for their potential impact on human communities. One of
the larger Holocene eruptions was that of Mt Mazama, which produced Crater
Lake (Klug et al., 2002). The main product of this eruption was highly
vesicular pumice in which most pores are connected. The high vesicle number
densities indicate that bubble nucleation occurred rapidly at a high degree of
supersaturation; hence continuous nucleation models cannot be applied. Small
vesicles are approximately lognormal by volume reflecting this single nucleation and growth event. Larger vesicles follow a power law distribution which
may be due to coalescence (Figure 7.5). Klug et al. (2002) proposed that
nucleation was in response to a downward propagating decompression
wave, followed by bubble growth and coalescence, and finally fragmentation.
Submarine pyroclastic volcanism seems at first glance to be unlikely, but
there is abundant evidence of its occurrence in some situations. Clearly, the
question involves degassing, hence the interest in evidence provided by VSDs
log (NV > L) (cm–3)
12
8
4
0
0
2
log (L = vesicle diameter) (μm)
4
Figure 7.5 Cumulative vesicle size distributions in pumice from Mt
Mazama (Klug et al., 2002). Size distributions of all samples depart from a
power law model (slope 3.3) at small sizes.
230
Textures of fluid-filled pores
measured elevation (km)
in pumice. The Healy caldera is an active volcano currently at a depth of
1.1–1.8 km. Wright et al. (2003) have quantified the shape and size of vesicles
in pumice from this eruption. There is a good correlation between vesicle size
and shape: increasing complexity with size is considered to reflect coalescence.
VSDs were measured by integrating slab and thin section data. Unfortunately,
the data were converted using inappropriate methods (see Section 3.3.4.6).
Despite this the VSD largely follows a power law between 0.03 and 2 mm with
d ¼ 2.5 0.4 (the effect of the conversion error on the VSD is unclear).
Departures from the distribution outside those limits are probably measurement artefacts. The VSD, magma water content and temperature indicate that
pyroclastic eruptions could have occurred at depths of 500–1000 m. As the
pyroclasts cooled they ingested water and sank.
The vesicle size distribution of lava flows has been proposed as a measure of
palaeoelevation of an eruption (Sahagian & Maus, 1994). The basis of this
method is that there is a difference in pressure between the base and top of the
flow caused by the mass of the lava. This pressure can be calculated and used
to calibrate the correlation between bubble sizes and pressure. The method
can only be applied to flows that have not been inflated after the bubbles
were frozen in at the base and top. The method was first applied to recent lavas
from Mauna Loa, Hawaii, and found to have a precision of about 400 m
(Figure 7.6; Sahagian et al., 2002a). It was subsequently applied to the problem
4.0
3.0
2.0
1.0
0
0
1.0
2.0
3.0
4.0
actual elevation (km)
Figure 7.6 Elevation determinations from differences in vesicle sizes
between the top and base of a flow against actual elevations (after Sahagian
et al., 2002a). These data are from recent lavas on the slopes of Mauna Loa,
Hawaii.
7.5 Typical applications
231
of the timing of uplift of the Colorado plateau during the last 23 Ma (Sahagian
et al., 2002b). Bubble size distributions indicated that the uplift rate was low
until 5 Ma ago when it increased by a factor of four.
Bubble shapes and orientations can also be used to infer flow type, shear
rate and shear stress in magmas (Rust et al., 2003). Bubbles in volcanic silicic
glasses are almost ideal strain-rate markers as they can be measured easily in
three dimensions and their relaxation times are long. The shape of bubbles is a
balance between shear forces that deform bubbles and surface tension (interfacial energy) that rerounds bubbles. Bubble shapes in three obsidian flows
were quantified in terms of bubble shape, size and orientation. These data
immediately provide data on the flow type: a spatter-fed flow deformed by
pure shear whereas simple shear dominated in juvenile obsidian from a pyroclastic deposit and a well-banded obsidian flow. Additional data on magma
viscosity and surface tension enabled the calculation of shear rates and shear
stresses.
Some glassy basaltic samples dredged from the ocean floor are so charged
with gas that they pop when brought to the surface. Burnard (1999) examined
one such sample that had 17% vesicles. He quantified the shape of the vesicles
against size and concluded that degassing occurred whilst the magma was
rising in a conduit at a speed of at least 1 ms1. This is comparable with
some explosive eruptions in Hawaii. Bubble size distributions were semilogarithmic, indicating that nucleation of bubbles was continuous and not a
single event.
Degassing and crystallisation are clearly linked in many volcanic systems,
but there have been few studies in which the two have been linked quantitatively. Gualda et al. (2004) have investigated quartz size distributions and total
vesicularity in the rhyolitic Bishop Tuff. They found that there was an inverse
correlation between the total vesicularity and the amount of the largest fraction of quartz crystals. They suggested that it was the result of the sinking of
large crystals and the rise of bubbles in the magma chamber. However, it is also
possible that the link is loss of gas during coarsening of the quartz in the
magma. Very long magma storage times are certainly consistent with this idea.
Notes
1. Fluid is used here in its physical sense for a gas, liquid or supercritical fluid. It therefore
includes fluids with a wide range in chemical composition, such as those dominated by water,
hydrocarbons, carbon dioxide, silicates, carbonates, sulphides, etc.
8
Appendix: Computer programs for use in
quantitative textural analysis (freeware,
shareware and commercial)
An updated version of these pages will be available at http://geologie.uqac.ca/
%7Emhiggins/CSD.html
8.1 Abbreviations
Cþþ
FORTRAN
IDL
Mac
MacX
Win
Linux
All
Excel
NIH
Free
Share
Com
Uncompiled program in Cþþ
Uncompiled program in FORTRAN
IDL language
Macintosh
Macintosh OS X
Windows
Linux
All of the above platforms
Excel worksheets and macros
NIHImage (ImageJ) macros
Freeware
Shareware
Commercial
8.2 Chapter 2: General analytical methods
This program can be used to segment, separate and measure data volumes
(3-D images).
BLOB3-D
IDL, All
ftp://ctlab.geo.utexas.edu/blob3d
Ketcham (2005)
These programs can be used to combine serial section images into a 3-D image.
232
233
Appendix
NIHImage
http://rsb.info.nih.gov/
nih-image/
ImageJ
http://rsb.info.nih.gov/ij/
Mac, Free
Win, Mac, MacX,
Linux, Free
This is a very good freeware
program. Many versions
are available
This is a java version of
NIHImage. It is being
upgraded and maintained
Bitmap images ( jpeg, tiff, bmp, etc.) can be stitched together with these
programs.
Adobe Photoshop
http://www.adobe.com/
ArcSoft Panorama maker
PanaVue ImageAssembler
http://www.panavue.com
Win, Mac, Com
Win, Mac, Com
Win, Com
The basic version is bundled with
some digital cameras
Bundled with some digital
cameras
Can stitch very large images
Images acquired by scanning photographs, rock blocks or thin sections
must be converted to binary images before they can be analysed automatically. There are a number of commercial and shareware programs available
including:
Adobe Photoshop
http://www.adobe.com/
Corel PhotoPaint
http://www.corel.com/
Win, Mac, Com
LViewPro
http://www.lview.com/
Win, Share
Image-Pro Plus
http://www.mediacy.com/
IPLab
http://www.iplab.com/sos/
product/gen/IPLab.html
Lazy grainboundaries
http://pages.unibas.ch/
earth/micro/index.html/
grainsize/grainsize.html
Win, Com
Win, Mac, Com
Mac, Win, Com
NIH, Free
This is the classic image
processing program
This program is very similar to
Photoshop and comes included
with Corel Draw
This program is quite compact
and does a lot for a shareware
program
This program is very good but
expensive
Commercial image analysis
program
Automatic tracing of grain
boundaries
234
Appendix
Binary images must be analysed to extract textural information. These
programs can do this.
Image-Pro Plus
http://www.mediacy.com/
IPLab
http://www.iplab.com/sos/
product/gen/IPLab.html
Aphelion
http://www.adcis.net
NIHImage
http://rsb.info.nih.gov/
nih-image/
ImageJ
http://rsb.info.nih.gov/ij/
ImageTool
http://ddsdx.uthscsa.edu/dig/
itdesc.html
Visilog
http://www.norpix.com/
visilog.htm
Win, Com
Mac, Win, Com
Win, Linux
Mac, Free
Win, Mac, MacX,
Linux, Free
Win, Free
Win, Com
This program is very good but
expensive
Commercial image analysis
program
Commercial image analysis
program
This is a very good shareware
program. Many versions are
available
This is a Java version of
NIHImage. It is being
upgraded and maintained
Freeware image processing,
originally developed for
biological work
Commercial image analysis
program
Digitising tablets and control programs.
Sigma-Scan
Win, Com
Made by Jandel Scientific.
http://www.spss.com/software/science/SigmaScan/
These special programs treats multiple images acquired at different orientations of the polarisers with computer-integrated polarising microscopy (CIP).
Geovision
http://
craton.geol.brocku.ca/
faculty/ff/stage/
WelcomeF.html
Image segmentation
ftp://ftp.iamg.org/VOL30/
v30-08-03.zip
Win, Free
Petrographic image processing of thin
sections using the rotating polariser
stage.
Linux, Free
Segmentation of petrographic images
acquired with CIP microscopy.
(Zhou et al., 2004b)
Appendix
235
Miscellaneous programs and sites.
MELTS
http://ctserver.
uchicago.edu/
MinIdent-Win
http://www.
micronex.ca/
Mac, Unix, Java
Win, Com
Calculation of liquid line of descent,
density and other parameters
(Ghiorso & Sack, 1995)
Mineral data database: mineral
properties include chemical
composition, optical, X-ray, physical,
hand-specimen. Data ranges compiled
from real analyses. About 6000 named
and unnamed minerals
International Society
for Stereology
http://www.stereology
society.org/
8.3 Chapter 3: Grain and crystal sizes
CSDCorrections
http://geologie.uqac.ca/
%7Emhiggins/
csdcorrections.html
Win, Free
CSD – Excel spreadsheet
(contact author)
StripStar
http://pages.unibas.ch/earth/
micro/index.html
Win, Free
Mac, Free
Distribution-free methods. Blocks,
ellipsoids, massive and oriented
textures. Log or linear size
intervals. Choice of output graphs.
Michael D Higgins, UQAC
Parametric solutions – Tony Peterson,
Geological Survey of Canada
Distribution-free methods. Conversion
of spheres only. Only uses fixed with
bins. Not user friendly. M Strip,
University of Basel
8.4 Chapter 4: Grain shape
ImageJ
http://rsb.info.
nih.gov/ij/
Fractscript
(contact author)
Win, Mac, MacX,
Linux, Free
Fraccalc
(contact author)
NIH
ImageTool macro
This is a Java version of NIHImage
Box-method fractal parameters can
be determined
A macro for calculating the fractal
dimension of object perimeters
in images of multiple objects
(Dillon et al., 2001)
M. Warfel in Dellino & Liotino (2002)
236
Appendix
Gaussian
DiffusionResample
(contact author)
NIH
(Saiki, 1997, Saiki et al., 2003)
8.5 Chapter 5: Grain orientations (rock fabric)
INTERCPT and SPO
http://www.sciences.
univ-nantes.fr/geol/
UMR6112/SPO/
index.html
Ellipsoid
http://www.sciences.
univ-nantes.fr/geol/
UMR6112/SPO/
index.html
Various programs
http://www.sp.uconn.edu/
geo102vc/Compaction_
Programs.htm
Win, Free
Program for applying the inertia
tensor and intercept method to
binary bitmap images. Patrick
Launeau and Pierre Robin
Win, Free
Ellipsoid calculation of 3-D shape
preferred orientation by
combination of 2-D images.
Patrick Launeau and Pierre Robin
Win, Excel,
Mathlab,
Free
Basel software package
http://www.unibas.ch/
earth/micro/software/
Lazy macros
http://www.unibas.ch/
earth/micro/software/
macros/lazymacros.html
Wavecalc
http://www.iamg.org/
CGEditor/cg1997.htm
Mac, NIH,
FORTRAN
Programs to estimate compaction
from scanned thin section images
and digitised plagioclase chain
networks. N. H. Gray and
A. Philpotts
Package of programs and macros for
computer-integrated polarisation
microscopy (CIP)
Package of programs and macros for
computer-integrated polarisation
microscopy (CIP)
NIH, Free
Cþþ
Software for multi-scale image
analysis: the normalised,
optimised anisotropic wavelet
coefficient method. Darrozes
(1997)
NOAWC pages
http://renass.u-strasbg.fr/
philippe/Couleur/
coul_titre.html
8.6 Chapter 6: Grain spatial distributions and relations
Big R calculations
http://geologie.uqac.ca/
%7Emhiggins/big_r.zip
Win
Michael D. Higgins, UQAC
References
Agrawal, Y., McCave, I. N. & Riley, J. (1991). Laser diffraction size analysis. In
J. P. M. Syvitski, ed., Principles, Methods, and Application of Particle Size
Analysis. New York: Cambridge University Press.
Akesson, U., Lindqvist, J. E., Goransson, M. & Stigh, J. (2001). Relationship between
texture and mechanical properties of granites, Central Sweden, by the use of
image-analysing technique. Bulletin of Engineering Geology and the Environment,
60, 277–84.
Akesson, U., Stigh, J., Lindqvist, J. E. & Goransson, M. (2003). The influence of
foliation on the fragility of granitic rocks, image analysis and quantitative
microscopy. Engineering Geology, 68, 275–88.
Allen, S. R. & McPhie, J. (2003). Phenocryst fragments in rhyolitic lavas and lava
domes. Journal of Volcanology and Geothermal Research, 126, 263–83.
Anderson, A. T. (1983). Oscillatory zoning of plagioclase: Nomarski interference
contrast microscopy of etched polished sections. American Mineralogist, 68, 125–9.
Arbaret, L., Fernandez, A., Jezek, J., Ildefonse, B., Launeau, P. & Diot, H. (2000).
Analogue and numerical modelling of shape fabrics: application to strain and
flow determination in magmas. Transactions of the Royal Society of
Edinburgh-Earth Sciences, 91, 97–109.
Armienti, P. & Tarquini, S. (2002). Power law olivine crystal size distributions in
lithospheric mantle xenoliths. Lithos, 65, 273–85.
Armienti, P., Pareschi, M. T., Innocenti, F. & Pompilio, M. (1994). Effects of magma
storage and ascent on the kinetics of crystal growth. The case of the 1991–93
Mt. Etna eruption. Contributions to Mineralogy and Petrology, 115, 402–14.
ASTM (1992). ASTM E930–92e1 Standard test methods for estimating the largest
grain observed in a metallographic section. Philadelphia, PA: American Society
for Testing Materials.
ASTM (1996). ASTM E112–96 Standard test methods for determining average grain
size. Philadelphia, PA: American Society for Testing Materials.
ASTM (1997). ASTM E1382–97 Standard test method for determining average grain
size using semiautomatic and automatic image analysis. Philadelphia, PA:
American Society for Testing Materials.
Baronnet, A. (1984). Growth kinetics of the silicates. A review of basic concepts.
Fortschritte der mineralogie, 62, 187–232.
Barrett, P. J. (1980). The shape of rock particles, a critical review. Sedimentology, 27,
291–303.
237
238
References
Bateman, P. C. & Chappell, B. W. (1979). Crystallisation, fractionation, and
solidification of the Tuolumne Intrusive series, Yosemite National Park,
California. Geological Society of America Bulletin, 90, 465–82.
Beere, W. (1975). Unifying theory of stability of penetrating liquid-phases and
sintering pores. Acta Metallurgica, 23, 131–8.
Ben Ismail, W. & Mainprice, D. (1998). An olivine fabric database: an
overview of upper mantle fabrics and seismic anisotropy. Tectonophysics,
296, 145–57.
Benn, K. & Mainprice, D. (1989). An interactive program for determination of
plagioclase crystal axes orientations from U-stage measurements – an aid for
petrofabric studies. Computers & Geosciences, 15, 1127–42.
Bennema, P., Meekes, H. & Enckevort, V. (1999). Crystal growth and morphology:
A multi-faceted approach. In B. Jamtveit & P. Meakin, eds., Growth, Dissolution,
and Pattern Formation in Geosystems. Dordrecht: Boston, pp. 21–64.
Berger, A. (2004). An improved equation for crystal size distribution in second-phase
influenced aggregates. American Mineralogist, 89, 126–31.
Berger, A. & Herwegh, M. (2004). Grain coarsening in contact metamorphic
carbonates: effects of second-phase particles, fluid flow and thermal
perturbations. Journal of Metamorphic Geology, 22, 459–74.
Berger, A. & Roselle, G. (2001). Crystallization processes in migmatites. American
Mineralogist, 86, 215–24.
Bevington, P. R. & Robinson, D. K. (2003). Data Reduction and Error Analysis for the
Physical Sciences. Boston, MA: McGraw-Hill.
Bindeman, I. & Valley, J. W. (2001). Low-18O rhyolites from Yellowstone:
Magmatic evolution based on analyses of zircons and individual phenocrysts.
Journal of Petrology, 42, 1491–517.
Bindeman, I. N. (2003). Crystal sizes in evolving silicic magma chambers. Geology,
31, 367–70.
Blanchard, J.-P., Boyer, P. & Gagny, C. (1979). Un nouveau critère de sens de mise en
place dans une caisse filonienne: Le ‘pincement’ des minéraux aux epontes
(Orientation des minéraux dans un magma en écoulement). Tectonophysics,
53, 1–25.
Blower, J. D., Keating, J. P., Mader, H. M. & Phillips, J. C. (2001). Inferring volcanic
degassing processes from vesicle size distributions. Geophysical Research Letters,
28, 347–50.
Blower, J. D., Keating, J. P., Mader, H. M. & Phillips, J. C. (2002). The evolution
of bubble size distributions in volcanic eruptions. Journal of Volcanology and
Geothermal Research, 120, 1–23.
Blumenfeld, P. & Bouchez, J. L. (1988). Shear criteria in granite and migmatite
deformed in the magmatic and solid states. Journal of Structural Geology, 10,
361–72.
Boorman, S., Boudreau, A. & Kruger, F. J. (2004). The lower zone-critical zone
transition of the Bushveld complex: A quantitative textural study. Journal of
Petrology, 45, 1209–35.
Borradaile, G. J. (1988). Magnetic susceptibility, petrofabrics and strain.
Tectonophysics, 156, 1–20.
Borradaile, G. J. & Gauthier, D. (2003). Interpreting anomalous magnetic fabrics in
ophiolite dikes. Journal of Structural Geology, 25, 171–82.
Borradaile, G. J. & Henry, B. (1997). Tectonic applications of magnetic susceptibility
and its anisotropy. Earth-Science Reviews, 42, 49–93.
References
239
Bouchez, J. L. (1997). Granite is never isotropic: an introduction to AMS studies of
granitic rocks. In J. L. Bouchez, D. H. W. Hutton & W. E. Stephens, eds.,
Granite: From Segregation of Melt to Emplacement Fabrics. Dordrecht: Kluwer
Academic Publishers.
Bouchez, J. L. (2000). Magnetic susceptibility anisotropy and fabrics in granites.
Comptes rendus de l’Academie des Sciences, 330, 1–14.
Boudreau, A. E. (1987). Pattern forming during crystallisation and the formation of
fine-scale layering. In I. Parsons, ed., Origins of Igneous Layering. Dordrecht:
D. Reidel, pp. 453–71.
Boudreau, A. E. (1995). Crystal aging and the formation of fine-scale igneous
layering. Mineralogy and Petrology, 54, 55–69.
Bowman, E. T., Soga, K. & Drummond, W. (2001). Particle shape characterisation
using Fourier descriptor analysis. Geotechnique, 51, 545–54.
Bozhilov, K. N., Green, H. W. & Dobrzhinetskaya, L. F. (2003). Quantitative 3D
measurement of ilmenite abundance in Alpe Arami olivine by confocal
microscopy: Confirmation of high-pressure origin. American Mineralogist,
88, 596–603.
Brandeis, G. & Jaupart, C. (1987). The kinetics of nucleation and crystal growth and
scaling laws for magmatic crystallisation. Contributions to Mineralogy and
Petrology, 96, 24–34.
Brandon, D. G. & Kaplan, W. D. (1999). Microstructural Characterization of
Materials. Chichester, NY: J. Wiley and Sons.
Brantley, S. L. & Mellott, N. P. (2000). Surface area and porosity of primary silicate
minerals. American Mineralogist, 85, 1767–83.
Brantley, S. L., White, A. F. & Hodson, M. E. (1999). Surface area of primary silicate
minerals. In B. Jamtveit & P. Meakin, eds., Growth, Dissolution, and Pattern
Formation in Geosystems. Dordrecht: Boston, pp. 291–326.
Bryon, D. N., Atherton, M. P. & Hunter, R. H. (1995). The interpretation of
granitic textures from serial thin sectioning, image-analysis and 3-dimensional
reconstruction. Mineralogical Magazine, 59, 203–11.
Bulau, J. R., Waff, H. S. & Tyburczy, J. A. (1979). Mechanical and thermodynamic
constraints on fluid distribution in partial melts. Journal of Geophysical Research,
84, 6102–8.
Bunge, H. J. (1982). Texture Analysis in Materials Science. London, UK:
Butterworths.
Burnard, P. (1999). Eruption dynamics of ‘popping rock’ from vesicle morphologies.
Journal of Volcanology and Geothermal Research, 92, 247–58.
Cabane, H., Laporte, D. & Provost, A. (2001). Experimental investigation of the
kinetics of Ostwald ripening of quartz in silicic melts. Contributions to Mineralogy
and Petrology, 142, 361–73.
Cabri, L. & Vaughan, D., eds., (1998). Modern Approaches to Ore and Environmental
Mineralogy. Short Course Series, 27. Ottawa: Mineralogical Association of
Canada.
Canon-Tapia, E. & Castro, J. (2004). AMS measurements on obsidian from the Inyo
Domes, CA: a comparison of magnetic and mineral preferred orientation fabrics.
Journal of Volcanology and Geothermal Research, 134, 169–82.
Canon-Tapia, E., Walker, G. P. L. & Herrero-Bervera, E. (1997). The internal
structure of lava flows – Insights from AMS measurements II: Hawaiian
pahoehoe, toothpaste lava and a’a. Journal of Volcanology and Geothermal
Research, 76, 19–46.
240
References
Capaccioni, B. & Sarocchi, D. (1996). Computer-assisted image analysis on clast
shape fabric from the Orvieto-Bagnoregio ignimbrite (Vulsini District, central
Italy): Implications on the emplacement mechanisms. Journal of Volcanology and
Geothermal Research, 70, 75–90.
Capaccioni, B., Valentini, L., Rocchi, M. B. L., Nappi, G. & Sarocchi, D. (1997).
Image analysis and circular statistics for shape-fabric analysis: Applications to
lithified ignimbrites. Bulletin of Volcanology, 58, 501–14.
Carey, S., Maria, A. & Sigurdsson, H. (2000). Use of fractal analysis for
discrimination of particles from primary and reworked jokulhlaup deposits in
SE Iceland. Journal of Volcanology and Geothermal Research, 104, 65–80.
Carlson, W. D. (1999). The case against Ostwald ripening of porphyroblasts.
Canadian Mineralogist, 37, 403–13.
Carlson, W. D., Denison, C. & Ketcham, R. A. (1995). Controls on the nucleation and
growth of porphyroblasts: Kinetics from natural textures and numerical models.
Geological Journal, 30, 207–25.
Cashman, K. (1986). Crystal size distributions in igneous and metamorphic rocks.
Baltimore, MA: Johns Hopkins University.
Cashman, K. & Blundy, J. (2000). Degassing and crystallization of ascending andesite
and dacite. Philosophical Transactions of the Royal Society of London Series A
(Mathematical Physical and Engineering Sciences), 358, 1487–513.
Cashman, K. V. (1988). Crystallisation of Mount St. Helens dacite; a quantitative
textural approach. Bulletin of Volcanology, 50, 194–209.
Cashman, K. V. (1990). Textural constraints on the kinetics of crystallization of
igneous rocks. In J. Nicholls & J. K. Russell, eds., Modern Methods of Igneous
Petrology: Understanding Magmatic Processes. Washington DC: Mineralogical
Society of America, pp. 259–314.
Cashman, K. V. (1992). Groundmass crystallisation of Mount St Helens dacite
1980–1986: a tool for interpreting shallow magmatic processes. Contributions
to Mineralogy and Petrology, 109, 431–49.
Cashman, K. V. (1993). Relationship between plagioclase crystallisation and cooling
rate in basaltic melts. Contributions to Mineralogy and Petrology, 113, 126–42.
Cashman, K. V. & Ferry, J. M. (1988). Crystal size distribution (CSD) in rocks and the
kinetics and dynamics of crystallization III. Metamorphic crystallization.
Contributions to Mineralogy and Petrology, 99, 410–15.
Cashman, K. V. & Marsh, B. D. (1988). Crystal size distribution (CSD) in rocks and
the kinetics and dynamics of crystallisation II. Makaopuhi lava lake.
Contributions to Mineralogy and Petrology, 99, 292–305.
Cashman, K. V., Thornber, C. & Kauahikaua, J. P. (1999). Cooling and
crystallization of lava in open channels, and the transition of pahoehoe lava
to a’a. Bulletin of Volcanology, 61, 306–23.
Castro, J. M., Cashman, K. V. & Manga, M. (2003). A technique for measuring 3D
crystal-size distributions of prismatic microlites in obsidian. American
Mineralogist, 88, 1230–40.
Chacron, M. & L’Heureux, I. (1999). A new model of periodic precipitation
incorporating nucleation, growth and ripening. Physics Letters A, 263, 70–7.
Chayes, F. (1950). On the bias of grain-size measurements made in thin section.
Journal of Geology, 58, 156–60.
Cheadle, M. J., Elliott, M. T. & McKenzie, D. (2004). Percolation threshold and
permeability of crystallizing igneous rocks: The importance of textural
equilibrium. Geology, 32, 757–60.
References
241
Christiansen, C. & Hartman, D. (1991). Principles, methods, and application of
particle size analysis. In J. P. M. Syvitski, ed., Principles, Methods, and
Application of Particle Size Analysis. New York: Cambridge University Press,
pp. 237–48.
Clark, A. H., Pearce, T. H., Roeder, P. L. & Wolfson, I. (1986). Oscillatory zoning
and other microstructures in magmatic olivine and augite; Nomarski interference
contrast observations on etched polished surfaces. American Mineralogist, 71
734–41.
Cloetens, P., Ludwig, W., Boller, E., Peyrin, F., Schlenker, M. & Baruchel, J. (2002).
3D imaging using coherent synchrotron radiation. Image Analysis and
Stereology, 21 (suppl. 1), S75–S86.
Cmiral, M., FitzGerald, J. D., Faul, U. H. & Green, D. H. (1998). A close look at
dihedral angles and melt geometry in olivine-basalt aggregates: a TEM study.
Contributions to Mineralogy and Petrology, 130, 336–45.
Coakley, J. & Syvitski, J. P. M. (1991). Sedigraph technique. In J. P. M. Syvitski, ed.,
Principles, Methods, and Application of Particle Size Analysis. New York:
Cambridge University Press, pp. 129–42.
Coogan, L. A., Thompson, G. & MacLeod, C. J. (2002). A textural and geochemical
investigation of high level gabbros from the Oman ophiolite: implications for the
role of the axial magma chamber at fast-spreading ridges. Lithos, 63, 67–82.
Costa, L. F. & Cesar, R. M. (2001). Shape Analysis and Classification: Theory and
Practice. Boca Raton, FL: CRC Press.
Craig, D. B. (1961). The Benford Plate. American Mineralogist, 46, 757–8.
Cressie, N. (1991). Statistics for Spatial Data. New York: Wiley-Interscience.
Cruden, A. R. & Launeau, P. (1994). Structure, magnetic fabric and emplacement of
the Archean Lebel stock, SW Abitibi greenstone belt. Journal of Structural
Geology, 16, 677–91.
Dana, J. D., Dana, E. S., Palache, C., Berman, H. M. & Frondel, C. (1944). The System
of Mineralogy of James Dwight Dana and Edward Salisbury Dana, Yale University,
1837–1892. New York: J. Wiley and Sons, London: Chapman and Hall.
Daniel, C. G. & Spear, F. S. (1999). The clustered nucleation and growth processes of
garnet in regional metamorphic rocks from north-west Connecticut, USA.
Journal of Metamorphic Geology, 17, 503–20.
Darrozes, J., Gaillot, P., De Saint-Blanquat, M. & Bouchez, J. L. (1997). Software
for multi-scale image analysis: The normalized optimized anisotropic wavelet
coefficient method. Computers & Geosciences, 23, 889–95.
Deakin, A. S. & Boxer, G. L. (1989). Argyle AK1 diamond size distribution; the use
of fine diamonds to predict the occurrence of commercial size diamonds. In
J. Ross, A. L. Jaques, J. Ferguson, D. H. Green, S. Y. O’Reilly, R. V. Danchin &
A. J. A. Janse, eds., Fourth International Kimberlite Conference. Sydney:
Geological Society of Australia, pp. 1117–22.
DeHoff, R. T. (1984). Generalized microstructural evolution by interface controlled
coarsening. Acta Metallurgica, 32, 43–7.
DeHoff, R. T. (1991). A geometrically general theory of diffusion controlled
coarsening. Acta Metallurgica et Materialia, 39, 2349–60.
Delesse, M. A. (1847). Procedé mécanique pour déterminer la composition des roches.
Comptes rendus de l’Acade´mie des Sciences (Paris), 25, 544–5.
Dellino, P. & Liotino, G. (2002). The fractal and multifractal dimension of volcanic
ash particles contour: a test study on the utility and volcanological relevance.
Journal of Volcanology and Geothermal Research, 113, 1–18.
242
References
Dillon, C. G., Carey, P. F. & Worden, R. H. (2001). Fractscript: A macro for
calculating the fractal dimension of object perimeters in images of multiple
objects. Computers & Geosciences, 27, 787–94.
Diot, H., Bolle, O., Lambert, J. M., Launeau, P. & Duchesne, J. C. (2003). The
Tellnes ilmenite deposit (Rogaland, South Norway): magnetic and petrofabric
evidence for emplacement of a Ti-enriched noritic crystal mush in a fracture zone.
Journal of Structural Geology, 25, 481–501.
Donaldson, C. H. (1976). An experimental investigation of olivine morphology.
Contributions to Mineralogy and Petrology, 57, 187–213.
Dowty, E. (1980). Crystal growth and nucleation theory and the numerical simulation
of igneous crystallization. In R. B. Hargraves, ed., Physics of Magmatic
Processes. Princeton: Princeton University Press.
Drolon, H., Hoyez, B., Druaux, F. & Faure, A. (2003). Multiscale roughness analysis
of particles: Application to the classification of detrital sediments. Mathematical
Geology, 35, 805–17.
Druitt, T. H., Edwards, L., Mellors, R. M., et al. (1999). Santorini Volcano. London,
UK: Geological Society. Memoir 19
Dunbar, N. W., Cashman, K. V. & Dupre, R. (1994). Crystallization processes of
anorthoclase phenocrysts in the Mount Erebus magmatic system; evidence from
crystal composition, crystal size distributions, and volatile contents of melt
inclusions. In P. R. Kyle, ed., Volcanological and Environmental Studies of
Mount Erebus, Antarctica. Washington DC: American Geophysical Union,
pp. 129–46.
Duyster, J. & Stockhert, B. (2001). Grain boundary energies in olivine derived
from natural microstructures. Contributions to Mineralogy and Petrology,
140, 567–76.
Eberl, D. D., Drits, V. A. & Srodon, J. (1998). Deducing growth mechanisms for
minerals from the shapes of crystal size distributions. American Journal of
Science, 298, 499–533.
Eberl, D. D., Kile, D. E. & Drits, V. A. (2002). On geological interpretations of crystal
size distributions: Constant vs. proportionate growth. American Mineralogist,
87, 1235–41.
Ehrlich, R. & Weinberg, B. (1970). An exact method for characterization of grain
shape. Journal of Sedimentary Petrology, 40, 205–12.
Eisenhour, D. D. (1996). Determining chondrule size distributions from thin-section
measurements. Meteoritics & Planetary Science, 31, 243–8.
Elliott, M. T. & Cheadle, M. J. (1997). On the identification of textural disequilibrium
in rocks using dihedral angle measurements: Reply. Geology, 25, 1055.
Elliott, M. T., Cheadle, M. J. & Jerram, D. A. (1997). On the identification of textural
equilibrium in rocks using dihedral angle measurements. Geology, 25, 355–8.
Emmons, R. C. (1964). The Universal Stage (With Five Axes of Rotation). New York:
Geological Society of America.
Epstein, B. (1947). The mathematical description of certain breakage mechanisms
leading to the logarithmico-normal distribution. Journal of the Franklin Institute,
244, 471–7.
Ernst, R. E. & Baragar, W. R. A. (1992). Evidence from magnetic fabric for the flow
pattern of magma in the Mackenzie giant radiating dyke swarm. Nature, 356,
511–13.
Ersoy, A. & Waller, M. D. (1995). Textural characterization of rocks. Engineering
Geology, 39, 123–36.
References
243
Exner, H. (2004). Stereology and 3D microscopy: Useful alternatives or competitors
in the quantitative analysis of microstructures? Image Analysis and Stereology,
23, 73–82.
Faul, U. (1997). The permeability of partially molten upper mantle rocks from
experiments and percolation theory. Journal of Geophysical Research, 102,
10299–311.
Faure, F., Trolliard, G., Nicollet, C. & Montel, J. M. (2003). A developmental
model of olivine morphology as a function of the cooling rate and the degree of
undercooling. Contributions to Mineralogy and Petrology, 145, 251–63.
Flinn, D. (1962). On folding during three-dimensional progressive deformation.
Geological Society of London Quarterly Journal, 118, 385–433.
Friedman, G. M. (1958). Determination of sieve-size distributions from thin-section
data for sedimentary petrological studies. Journal of Geology, 66, 394–416.
Fueten, F. (1997). A computer-controlled rotating polarizer stage for the petrographic
microscope. Computers & Geosciences, 23, 203–8.
Fueten, F. & Goodchild, J. S. (2001). Quartz c-axes orientation determination using
the rotating polarizer microscope. Journal of Structural Geology, 23, 895–902.
Gaillot, P., Darrozes, J., de Saint Blanquat, M. & Ouillon, G. (1997). The normalized
optimized anisotropic wavelet coefficient (NOAWC) method; an image
processing tool for multiscale analysis of rock fabric. Geophysical Research
Letters, 24, 1819–22.
Gaillot, P., Darrozes, J. & Bouchez, J. L. (1999). Wavelet transform: a future of rock
fabric analysis? Journal of Structural Geology, 21, 1615–21.
Galwey, A. K. & Jones, K. A. (1963). An attempt to determine the mechanism of a
natural mineral-forming reaction from examination of the products. Journal of
the Chemical Society (London), 5681–6.
Galwey, A. K. & Jones, K. A. (1966). Crystal size frequency distribution of garnets
in some analysed metamorphic rocks from Mallaig, Inverness, Scotland.
Geological Magazine, 103, 143–52.
Gaonac’h, H., Lovejoy, S., Stix, J. & Scherzter, D. (1996a). A scaling growth model
for bubbles in basaltic lava flows. Earth and Planetary Science Letters, 139,
395–409.
Gaonac’h, H., Stix, J. & Lovejoy, S. (1996b). Scaling effects on vesicle shape, size
and heterogeneity of lavas from Mount Etna. Journal of Volcanology and
Geothermal Research, 74, 131–52.
Gaonac’h, H., Lovejoy, S. & Schertzer, D. (2003). Percolating magmas and explosive
volcanism. Geophysical Research Letters, 30, 1559.
Gardner, J. E. & Denis, M.-H. (2004). Heterogeneous bubble nucleation on Fe-Ti
oxide crystals in high-silica rhyolitic melts. Geochimica et Cosmochimica Acta,
68, 3587–97.
Gardner, J. E., Hilton, M. & Carroll, M. R. (1999). Experimental constraints on
degassing of magma: isothermal bubble growth during continuous decompression
from high pressure. Earth and Planetary Science Letters, 168, 201–18.
Garrido, C. J., Kelemen, P. B. & Hirth, G. (2001). Variation of cooling rate with depth
in lower crust formed at an oceanic spreading ridge; plagioclase crystal size
distributions in gabbros from the Oman Ophiolite. Geochemistry Geophysics
Geosystems, doi: 10.1029/2000GC000136.
Gee, J. S., Meurer, W. P., Selkin, P. A. & Cheadle, M. J. (2004). Quantifying threedimensional silicate fabrics in cumulates using cumulative distribution functions.
Journal of Petrology, 45, 1983–2009.
244
References
Geoffroy, L., Callot, J. P., Aubourg, C. & Moreira, M. (2002). Magnetic and
plagioclase linear fabric discrepancy in dykes: a new way to define the flow vector
using magnetic foliation. Terra Nova, 14, 183–90.
Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass-transfer in magmatic processes.
4. A revised and internally consistent thermodynamic model for the interpolation
and extrapolation of liquid-solid equilibria in magmatic systems at elevatedtemperatures and pressures. Contributions to Mineralogy and Petrology, 119,
197–212.
Gingras, M. K., MacMillan, B. & Balcom, B. J. (2002). Visualizing the internal
physical characteristics of carbonate sediments with magnetic resonance imaging
and petrography. Bulletin of Canadian Petroleum Geology, 50, 363–9.
Goodchild, J. S. & Fueten, F. (1998). Edge detection in petrographic images using the
rotating polarizer stage. Computers & Geosciences, 24, 745–51.
Goodrich, C. A. (2003). Petrogenesis of olivine-phyric shergottites Sayh Al Uhaymir
005 and elephant moraine A79001 lithology A. Geochimica et Cosmochimica
Acta, 67, 3735–72.
Gray, N. H. (1970). Crystal growth and nucleation in two large diabase dykes.
Canadian Journal of Earth Sciences, 7, 366–75.
Gray, N. H., Philpotts, A. R. & Dickson, L. D. (2003). Quantitative measures of
textural anisotropy resulting from magmatic compaction illustrated by a sample
from the Palisades sill, New Jersey. Journal of Volcanology and Geothermal
Research, 121, 293–312.
Gregg, S. J. & Sing, K. S. W. (1982). Adsorption, Surface Area, and Porosity. London:
Academic Press.
Gregoire, V., Darrozes, J., Gaillot, P., Nedelec, A. & Launeau, P. (1998). Magnetite
grain shape fabric and distribution anisotropy vs rock magnetic fabric: a threedimensional case study. Journal of Structural Geology, 20, 937–44.
Greshake, A., Fritz, J. & Stoffler, D. (2004). Petrology and shock metamorphism of
the olivine-phyric shergottite Yamato 980459: Evidence for a two-stage cooling
and a single-stage ejection history. Geochimica et Cosmochimica Acta, 68,
2359–77.
Gualda, G., Cook, D., Chopra, R. et al. (2004). Fragmentation, nucleation
and migration of crystals and bubbles in the Bishop Tuff rhyolitic
magma. Transactions of the Royal Society of Edinburgh-Earth Sciences,
95, 375–90.
Habesch, S. M. (2000). Electron backscattered diffraction analyses combined with
environmental scanning electron microscopy: potential applications for nonconducting, uncoated mineralogical samples. Materials Science and Technology,
16, 1393–8.
Hammer, J. E., Cashman, K. V., Hoblitt, R. P. & Newman, S. (1999). Degassing
and microlite crystallization during pre-climactic events of the 1991 eruption of
Mt. Pinatubo, Philippines. Bulletin of Volcanology, 60, 355–80.
Hanchar, J. M. & Hoskin, P. W. O., eds., (2003). Zircon. Reviews in Mineralogy and
Geochemistry, 53. Washington, DC: Mineralogical Society of America.
Harvey, P. K. & Laxton, R. R. (1980). The estimate of finite strain from the
orientation distribution of passively deformed linear markers: eigenvalue
relationships. Tectonophysics, 70, 285–307.
Heilbronner, P. R. & Bruhn, D. (1998). The influence of three-dimensional grain
size distributions on the rheology of polyphase rocks. Journal of Structural
Geology, 20, 695–705.
References
245
Heilbronner, R. (2002). Analysis of bulk fabrics and microstructure variations
using tesselations of autocorrelation functions. Computers & Geosciences,
28, 447–55.
Heilbronner, R. P. (1992). The autocorrelation function – an image-processing tool
for fabric analysis. Tectonophysics, 212, 351–70.
Heilbronner, R. P. & Pauli, C. (1993). Integrated spatial and orientation analysis
of quartz c-axes by computer-aided microscopy. Journal of Structural Geology,
15, 369–82.
Herring, C. (1951a). Some theorems on the free energies of crystal surfaces. Physics
Reviews, 82, 87–93.
Herring, C. (1951b). Surface tension as a motivation for sintering. In W. Kingston,
ed., Physics of Powder Metallurgy. New York: McGraw-Hill.
Herwegh, M. (2000). A new technique to automatically quantify microstructures of
fine grained carbonate mylonites: two-step etching combined with SEM imaging
and image analysis. Journal of Structural Geology, 22, 391–400.
Herwegh, M., de Bresser, J. & ter Heege, J. (2005). Combining natural microstructures
with composite flow laws: an improved approach for the extrapolation of lab
data to nature. Journal of Structural Geology, 27, 503–21.
Hext, G. (1963). The estimation of second-order tensors, with related tests and
designs. Biometrika, 50, 353–73.
Heyraud, J. C. & Metois, J. J. (1987). Equilibrium shape of an ionic-crystal in
equilibrium with its vapor (NaCl). Journal of Crystal Growth, 84, 503–8.
Higgins, M. D. (1991). The origin of laminated and massive anorthosite, Sept Iles
intrusion, Quebec, Canada. Contributions to Mineralogy and Petrology, 106,
340–54.
Higgins, M. D. (1994). Determination of crystal morphology and size from bulk
measurements on thin sections: numerical modelling. American Mineralogist,
79, 113–19.
Higgins, M. D. (1996a). Crystal size distributions and other quantitative textural
measurements in lavas and tuff from Mt Taranaki (Egmont volcano), New
Zealand. Bulletin of Volcanology, 58, 194–204.
Higgins, M. D. (1996b). Magma dynamics beneath Kameni volcano, Greece, as
revealed by crystal size and shape measurements. Journal of Volcanology and
Geothermal Research, 70, 37–48.
Higgins, M. D. (1998). Origin of anorthosite by textural coarsening: Quantitative
measurements of a natural sequence of textural development. Journal of
Petrology, 39, 1307–25.
Higgins, M. D. (1999). Origin of megacrysts in granitoids by textural coarsening: a
crystal size distribution (CSD) study of microcline in the Cathedral Peak
granodiorite, Sierra Nevada, California. In C. Fernandez & A. Castro, eds.,
Understanding Granites: Integrating Modern and Classical Techniques. Special
Publication 158. London: Geological Society of London, pp. 207–19.
Higgins, M. D. (2000). Measurement of crystal size distributions. American
Mineralogist, 85, 1105–16.
Higgins, M. D. (2002a). Closure in crystal size distributions (CSD), verification
of CSD calculations, and the significance of CSD fans. American Mineralogist,
87, 171–5.
Higgins, M. D. (2002b). A crystal size-distribution study of the Kiglapait layered
mafic intrusion, Labrador, Canada: evidence for textural coarsening.
Contributions to Mineralogy and Petrology, 144, 314–30.
246
References
Higgins, M. D. (2005). A new model for the structure of the Sept Iles intrusive suite,
Canada. Lithos, 83, 199–213.
Higgins, M. D. (2006). Use of appropriate diagrams to determine if crystal size
distributions (CSD) are dominantly semi-logarithmic, lognormal or fractal (scale
invariant). Journal of Volcanology and Geothermal Research.
Higgins, M. D. & Roberge, J. (2003). Crystal size distribution (CSD) of plagioclase
and amphibole from Soufriere Hills volcano, Montserrat: Evidence for dynamic
crystallisation / textural coarsening cycles. Journal of Petrology, 44, 1401–11.
Hiraga, T., Nishikawa, O., Nagase, T. & Akizuki, M. (2001). Morphology of
intergranular pores and wetting angles in pelitic schists studied by transmission
electron microscopy. Contributions to Mineralogy and Petrology, 141, 613–22.
Hiraga, T., Nishikawa, O., Nagase, T., Akizuki, M. & Kohlstedt, D. L. (2002).
Interfacial energies for quartz and albite in pelitic schist. Contributions to
Mineralogy and Petrology, 143, 664–72.
Holness, M. B. (1993). Temperature and pressure-dependence of quartz aqueous fluid
dihedral angles – the control of adsorbed H2O on the permeability of quartzites.
Earth and Planetary Science Letters, 117, 363–77.
Holness, M. B. (2005). Spatial constraints on magma chamber replenishment events
from textural observations of cumulates: the Rum layered intrusion, Scotland.
Journal of Petrology, 46, 1585–601.
Holness, M. B. & Siklos, S. T. C. (2000). The rates and extent of textural equilibration
in high-temperature fluid-bearing systems. Chemical Geology, 162, 137–53.
Holness, M. B., Cheadle, M. C. & McKenzie, D. (2005). On the use of changes in
dihedral angle to decode late-stage textural evolution in cumulates. Journal of
Petrology, 46, 1565–83.
Houghton, B. F. & Wilson, C. J. N. (1989). A vesicularity index for pyroclastic
deposits. Bulletin of Volcanology, 51, 451–62.
Howard, V. & Reed, M. G. (1998). Unbiased Stereology: Three-Dimensional
Measurement in Microscopy. Oxford, UK: Bios Scientific Publishers, New York:
Springer.
Howarth, D. & Rowlands, J. (1986). Development of an index to quantify rock
textures for quantitative assessment of intact rock properties. Geotechnical
Testing Journal, 9, 169–79.
Hunter, R. H. (1987). Textural Equilibrium in Layered Igneous Rocks. In I. Parsons,
ed., Origins of Igneous Layering. Dordrecht: D. Reidel, pp. 473–503.
Hunter, R. H. (1996). Textural Development in Cumulate Rocks. In R. G. Cawthorn,
ed., Layered Intrusions. Amsterdam: Elsevier, pp. 77–101.
Hurwitz, S. & Navon, O. (1994). Bubble nucleation in rhyolitic melts – experiments
at high-pressure, temperature, and water-content. Earth and Planetary Science
Letters, 122, 267–80.
Hutchison, C. S. (1974). Laboratory Handbook of Petrographic Techniques. Hoboken
NY: John Wiley & Sons.
Iezzi, G. & Ventura, G. (2002). Crystal fabric evolution in lava flows: results from
numerical simulations. Earth and Planetary Science Letters, 200, 33–46.
Ihinger, P. D. & Zink, S. I. (2000). Determination of relative growth rates of natural
quartz crystals. Nature, 404, 865–9.
Ikeda, S., Toriumi, M., Yoshida, H. & Shimizu, I. (2002). Experimental study of the
textural development of igneous rocks in the late stage of crystallization: the
importance of interfacial energies under non-equilibrium conditions.
Contributions to Mineralogy and Petrology, 142, 397–415.
References
247
Ikeda, S., Nakano, T., Tsuchiyama, A., Uesugi, K., Suzuki, Y., Nakamura, K.,
Nakashima, Y. & Yoshida, H. (2004). Nondestructive three-dimensional elementconcentration mapping of a Cs-doped partially molten granite by X-ray computed
tomography using synchrotron radiation. American Mineralogist, 89, 1304–13.
Ildefonse, B., Launeau, P. & Bouchez, J.-L. (1992). Effect of mechanical interactions
on the development of shape preferred orientations: a two-dimensional
experimental approach. Journal of Structural Geology, 14, 73–83.
Jackson, E. D. (1961). Primary textures and mineral associations in the ultramafic
zone of the Stillwater complex, Montana. United States Geological Survey
Professional Paper, 358.
Jackson, M. (1991). Anisotropy of magnetic remanence – a brief review of mineralogical
sources, physical origins, and geological applications, and comparison with
susceptibility anisotropy. Pure and Applied Geophysics, 136, 1–28.
Jackson, M., Gruber, W., Marvin, J. & Banerjee, S. K. (1988). Partial anhysteretic
remanence and its anisotropy – applications and grainsize-dependence.
Geophysical Research Letters, 15, 440–3.
Jamtveit, B. & Meakin, P., eds., (1999). Growth, Dissolution, and Pattern Formation in
Geosystems. Dordrecht: Kluwer Academic Publishers.
Jelinek, V. (1978). Statistical processing of anisotropy of magneticsusceptibility measured on groups of specimens. Studia Geophysica et
Geodaetica, 22, 50–62.
Jerram, D. A. (2001). Visual comparators for degree of grain-size sorting in two and
three-dimensions. Computers & Geoscience, 27, 485–92.
Jerram, D. A. & Cheadle, M. J. (2000). On the cluster analysis of grains and crystals in
rocks. American Mineralogist, 85, 47–67.
Jerram, D. A., Cheadle, M. J., Hunter, R. H. & Elliott, M. T. (1996). The spatial
distribution of grains and crystals in rocks. Contributions to Mineralogy and
Petrology, 125, 60–74.
Jerram, D. A., Cheadle, M. J. & Philpotts, A. R. (2003). Quantifying the building
blocks of igneous rocks: Are clustered crystal frameworks the foundation?
Journal of Petrology, 44, 2033–51.
Ji, S. C., Zhao, X. O. & Zhao, P. L. (1994). On the measurement of plagioclase lattice
preferred orientations. Journal of Structural Geology, 16, 1711–18.
Jillavenkateas, A., Dapkunas, S. J. & Lum, L.-S. H. (2001). Particle Size
Characterisation: NIST Recommended Practice Guide. Washington DC, USA:
National Institute of Standards and Technology.
Johnson, M. R. (1994). Thin-section grain-size analysis revisited. Sedimentology, 41,
985–99.
Jung, H. & Karato, S. (2001). Effects of water on dynamically recrystallized grain-size
of olivine. Journal of Structural Geology, 23, 1337–44.
Jurewicz, S. R. & Jurewicz, A. J. G. (1986). Distribution of apparent angles on random
sections with emphasis on dihedral angle measurements. Journal of Geophysical
Research-Solid Earth and Planets, 91, 9277–82.
Jurewicz, S. R. & Watson, E. B. (1985). The distribution of partial melt in a granitic
system – the application of liquid-phase sintering theory. Geochimica et
Cosmochimica Acta, 49, 1109–21.
Karato, S. & Wenk, H.-R. (2002). Plastic Deformation of Minerals and Rocks.
Washington, DC: Mineralogical Society of America.
Ketcham, R. A. (2005). Computational methods for quantitative analysis of threedimensional features in geological specimens. Geosphere, 1, 32–41.
248
References
Ketcham, R. A. & Carlson, W. D. (2001). Acquisition, optimization and
interpretation of X-ray computed tomographic imagery; applications to the
geosciences. Computers & Geoscience, 27, 381–400.
Ketcham, R. A. & Ryan, T. M. (2004). Quantification and visualization of anisotropy
in trabecular bone. Journal of Microscopy-Oxford, 213, 158–71.
Kile, D. E., Eberl, D. D., Hoch, A. R. & Reddy, M. M. (2000). An assessment of
calcite crystal growth mechanisms based on crystal size distributions. Geochimica
et Cosmochimica Acta, 64, 2937–50.
Klug, C. & Cashman, K. (1994). Vesiculation of May 18, 1980 Mount St. Helens
magma. Geology, 22, 468–72.
Klug, C., Cashman, K. V. & Bacon, C. R. (2002). Structure and physical
characteristics of pumice from the climactic eruption of Mount Mazama (Crater
Lake), Oregon. Bulletin of Volcanology, 64, 486–501.
Knight, M. D., Walker, G. P. L., Ellwood, B. B. & Diehl, J. F. (1986). Stratigraphy,
paleomagnetism, and magnetic fabric of the Toba Tuffs – constraints on the
sources and eruptive styles. Journal of Geophysical Research-Solid Earth and
Planets, 91, 355–82.
Kocks, U. F., Wenk, H.-R. & Tomé, C. N. (2000). Texture and Anisotropy: Preferred
Orientations in Polycrystals and Their Effect on Material Properties. Cambridge,
UK: Cambridge University Press.
Kolmogorov, A. (1941). The lognormal law of distribution of particle sizes during
crushing. Doklady Akademii Nauk SSSR, 31, 99–101.
Kong, M. Y., Bhattacharya, R. N., James, C. & Basu, A. (2005). A statistical
approach to estimate the 3D size distribution of spheres from 2D size
distributions. Geological Society of America Bulletin, 117, 244–9.
Kostov, I. & Kostov, R. I. (1999). Crystal Habits of Minerals. Sofia: Prof. Marin
Drinov Academic Publishing House; Pensoft Publishers.
Kotov, S. & Berendsen, P. (2002). Statistical characteristics of xenoliths in the Antioch
kimberlite pipe, Marshall county, northeastern Kansas. Natural Resources
Research, 11, 289–97.
Kouchi, A., Tsuchiyama, A. & Sunagawa, I. (1986). Effects of stirring on
crystallization of basalt: Texture and element partitioning. Contributions to
Mineralogy and Petrology, 93, 429–38.
Kretz, R. (1966a). Grain-size distribution for certain metamorphic minerals in
relation to nucleation and growth. Journal of Geology, 74, 147–73.
Kretz, R. (1966b). Interpretation of shape of mineral grains in metamorphic rocks.
Journal of Petrology, 7, 68–94.
Kretz, R. (1969). On the spatial distribution of crystals in rocks. Lithos, 2, 39–65.
Kretz, R. (1993). A garnet population in Yellowknife schist, Canada. Journal of
Metamorphic Geology, 11, 101–20.
Krug, H. J., Brandtstadter, H. & Jacob, K. H. (1996). Morphological instabilities in
pattern formation by precipitation and crystallization processes. Geologische
Rundschau, 85, 19–28.
Kruhl, J. H. & Nega, M. (1996). The fractal shape of sutured quartz grain boundaries:
Application as a geothermometer. Geologische Rundschau, 85, 38–43.
Kruhl, J. H. & Peternell, M. (2002). The equilibration of high-angle grain boundaries
in dynamically recrystallized quartz: the effect of crystallography and
temperature. Journal of Structural Geology, 24, 1125–37.
Lane, A. C. (1898). Geological Report on Isle Royale, Michigan. Lansing, MI:
Michigan Geological Survey.
References
249
Laporte, D. & Provost, A. (2000). Equilibrium geometry of a fluid phase in a
polycrystalline aggregate with anisotropic surface energies: Dry grain
boundaries. Journal of Geophysical Research-Solid Earth, 105, 25937–53.
Laporte, D. & Watson, E. B. (1995). Experimental and theoretical constraints on melt
distribution in crustal sources – the effect of crystalline anisotropy on melt
interconnectivity. Chemical Geology, 124, 161–84.
Larsen, J. F. & Gardner, J. E. (2000). Experimental constraints on bubble interactions
in rhyolite melts; implications for vesicle size distributions. Earth and Planetary
Science Letters, 180, 201–14.
Larsen, J. F., Denis, M. H. & Gardner, J. E. (2004). Experimental study of bubble
coalescence in rhyolitic and phonolitic melts. Geochimica et Cosmochimica Acta,
68, 333–44.
Larsen, L. & Poldervaart, A. (1957). Measurements and distribution of zircons in
some granitic rocks of magmatic origin. Mineralogical Magazine, 31, 544–64.
Lasaga, A. C. (1998). Kinetic Theory in the Earth Sciences. Princeton NJ: Princeton
University Press.
Launeau, P. (2004). Mise en evidence des écoulments magmatiques par analyse
d’images 2-D des distibutions 3-D d’orientations préférentielles de formes.
Bulletin de la Socie´te´ Ge´ologique de France, 175, 331–50.
Launeau, P. & Cruden, A. R. (1998). Magmatic fabric acquisition mechanisms in a
syenite: results of a combined AMS and image analysis study. Journal of
Geophysical Research, 103, 5067–89.
Launeau, P. & Robin, P. Y. F. (1996). Fabric analysis using the intercept method.
Tectonophysics, 267, 91–119.
Launeau, P., Bouchez, J. L. & Benn, K. (1990). Shape preferred orientation
of object populations; automatic analysis of digitized images. Tectonophysics,
180, 201–11.
Launeau, P., Cruden, A. R. & Bouchez, J. L. (1994). Mineral recognition in digital
images of rocks – a new approach using multichannel classification. Canadian
Mineralogist, 32, 919–33.
Lemelle, L., Simionovici, A., Truche, R. et al. (2004). A new nondestructive X-ray
method for the determination of the 3D mineralogy at the micrometer scale.
American Mineralogist, 89, 547–53.
Lentz, R. C. F. & McSween, H. Y., Jr. (2000). Crystallization of the basaltic
shergottites; insights from crystal size distribution (CSD) analysis of pyroxenes.
Meteoritics & Planetary Science, 35, 919–27.
Lewis, D. W. & McConchie, D. (1994a). Analytical Sedimentology. New York:
Chapman & Hall.
Lewis, D. W. & McConchie, D. (1994b). Practical Sedimentology. New York:
Chapman & Hall.
Lifshitz, I. M. & Slyozov, V. V. (1961). The kinetics of precipitation from supersaturated
solid solutions. Journal of Physics and Chemistry of Solids, 19, 35–50.
Lofgren, G. E. (1974). An experimental study of plagioclase crystal morphology:
isothermal crystallization. American Journal of Science, 274, 243–73.
Lofgren, G. E. & Donaldson, C. H. (1975). Curved branching crystals and
differentiation in comb-layered rocks. Contributions to Mineralogy and
Petrology, 49, 309–19.
Maaloe, S., Tumyr, O. & James, D. (1989). Population density and zoning of olivine
phenocrysts in tholeiites from Kauai, Hawaii. Contributions to Mineralogy and
Petrology, 101, 176–86.
250
References
Mainprice, D. & Nicolas, A. (1989). Development of shape and lattice preferred
orientations – application to the seismic anisotropy of the lower crust. Journal of
Structural Geology, 11, 175–89.
Mandelbrot, B. B. (1982). The Fractal Geometry of Nature. San Francisco: W. H.
Freeman.
Manga, M. (1998). Orientation distribution of microlites in obsidian. Journal of
Volcanology and Geothermal Research, 86, 107–15.
Mangan, M. T. (1990). Crystal size distribution and the determination of magma
storage times: The 1959 eruption of Kilauea volcano, Hawaii. Journal of
Volcanology and Geothermal Research, 44, 295–302.
Mangan, M. T. & Cashman, K. V. (1996). The structure of basaltic scoria and
reticulite and inferences for vesiculation, foam formation, and fragmentation in
lava fountains. Journal of Volcanology and Geothermal Research, 73, 1–18.
Mangan, M. T., Cashman, K. V. & Newman, S. (1993). Vesiculation of basaltic
magma during eruption. Geology, 21, 157–60.
Mardia, K. V. & Jupp, P. E. (2000). Directional Statistics. Chichester; New York:
John Wiley & Sons.
Markov, I. V. (1995). Crystal Growth for Beginners: Fundamentals of Nucleation,
Crystal Growth, and Epitaxy. Singapore: River Edge N.J.
Marqusee, J. A. & Ross, J. (1983). Kinetics of phase-transitions – theory of Ostwald
ripening. Journal of Chemical Physics, 79, 373–8.
Marschallinger, R. (1997). Automatic mineral classification in the macroscopic scale.
Computers & Geosciences, 23, 119–26.
Marschallinger, R. (1998a). Correction of geometric errors associated with the 3-D
reconstruction of geological materials by precision serial lapping. Mineralogical
Magazine, 62, 783–92.
Marschallinger, R. (1998b). A method for three-dimensional reconstruction of
macroscopic features in geological materials. Computers & Geosciences, 24, 875–83.
Marschallinger, R. (2001). Three-dimensional reconstruction and visualization of
geological materials with IDL – examples and source code. Computers &
Geosciences, 27, 419–26.
Marsh, B. D. (1988a). Crystal capture, sorting, and retention in convecting magma.
Geological Society of America Bulletin, 100, 1720–37.
Marsh, B. D. (1988b). Crystal size distribution (CSD) in rocks and the kinetics and
dynamics of crystallization I. Theory. Contributions to Mineralogy and Petrology,
99, 277–91.
Marsh, B. D. (1998). On the interpretation of crystal size distributions in magmatic
systems. Journal of Petrology, 39, 553–600.
Martin-Hernandez, F., Luneburg, C., Aubourg, C. & Jackson, M., eds., (2005).
Magnetic Fabric: Methods and Applications. Geological Society Special
Publication, 238. London: The Geological Society.
McBirney, A. R. & Hunter, R. H. (1995). The cumulate paradigm reconsidered.
Journal of Geology, 103, 114–22.
McBirney, A. R. & Nicolas, A. (1997). The Skaergaard layered series: Part II.
Dynamic layering. Journal of Petrology, 38, 569–80.
McConnell, J. (1975). Microstructures of minerals as petrogenetic indictors. Annual
Review of Earth and Planetary Sciences, 3, 129–55.
McKay, D. S., Gibson, E. K., Thomas-Keprta, K. L. et al. (1996). Search for past life
on Mars: Possible relic biogenic activity in Martian meteorite ALH84001.
Science, 273, 924–30.
References
251
McKenzie, D. (1984). The generation and compaction of partially molten rock.
Journal of Petrology, 25, 713–65.
Medley, E. W. (2002). Estimating block size distributions of melanges and similar
block-in-matrix rocks (bimrocks). In R. Hammah, W. Bawden, J. Curran &
M. Telesnicki, eds., North American Rock Mechanics Symposium. Toronto,
Canada: University of Toronto Press, pp. 509–606.
Mees, F., Swennen, R., Van Geet, M. & Jacobs, P., eds., (2003). Applications of X-ray
Computed Tomography in the Geosciences. Geological Society Special
Publication, 215. London: The Geological Society.
Meng, B. (1996). Determination and interpretation of fractal properties of the
sandstone pore system. Materials and Structures, 29, 195–205.
Merriam, D. F. (2004). The quantification of geology: from abacus to Pentium:
A chronicle of people, places, and phenomena. Earth-Science Reviews, 67, 55–89.
Meurer, W. P. & Boudreau, A. E. (1998). Compaction of igneous cumulates; Part II,
Compaction and the development of igneous foliations. Journal of Geology,
106, 293–304.
Middleton, G. V. (2000). Data Analysis in the Earth Sciences Using MATLAB1.
Upper Saddle River, NJ: Prentice Hall.
Milligan, T. & Kranck, K. (1991). Electroresistance particle size analysers. In
J. P. M. Syvitski, ed., Principles, Methods, and Application of Particle Size
Analysis. New York: Cambridge University Press.
Miyazaki, K. (1996). A numerical simulation of textural evolution due to Ostwald
ripening in metamorphic rocks: A case for small amount of volume of dispersed
crystals. Geochimica et Cosmochimica Acta, 60, 277–90.
Miyazaki, K. (2000). The case against Ostwald ripening of porphyroblasts:
Discussion. Canadian Mineralogist, 38, 1027–8.
Mock, A., Jerram, D. A. & Breitkreuz, C. (2003). Using quantitative textural analysis
to understand the emplacement of shallow-level rhyolitic laccoliths – A case study
from the Halle Volcanic Complex, Germany. Journal of Petrology, 44, 833–49.
Mora, C. F. & Kwan, A. K. H. (2000). Sphericity, shape factor, and convexity
measurement of coarse aggregate for concrete using digital image processing.
Cement and Concrete Research, 30, 351–8.
Morishita, R. (1998). Statistical properties of ideal rock textures: Relationship
between crystal size distribution and spatial correlation in minerals.
Mathematical Geology, 30, 409–34.
Morishita, R. & Obata, M. (1995). A new statistical description of the spatial
distribution of minerals in rocks. Journal of Geology, 103, 232–40.
Morse, S. A. (1969). The Kiglapait layered intrusion, Labrador. Geological Society of
America, Memoir, 112, 204.
Morse, S. A. (1979). Kiglapait geochemistry, II. Petrography. Journal of Petrology,
20, 591–624.
Muir, I. D. (1981). The 4-Axis Universal Stage. Chicago: Microscope Publications.
Mungall, J. & Su, S. (2005). Interfacial tension between sulfide and silicate liquids:
Constraints on kinetics of sulfide liquation and sulfide migration through silicate
rocks. Earth and Planetary Science Letters, 234, 135–49.
Murthy, D. N. P., Xie, M. & Jiang, R. (2004). Weibull Models. Wiley Series in
Probability and Statistics. Hoboken NJ: John Wiley and Sons.
Naslund, H. R. & McBirney, A. R. (1996). Mechanisms of formation of igneous
layering. In R. G. Cawthorn, ed., Layered Intrusions. Amsterdam: Elsevier,
pp. 1–44.
252
References
Nemchin, A. A., Giannini, L. M., Bodorkos, S. & Olivier, N. H. S. (2001). Ostwald
ripening as a possible mechanism for zircon overgrowth formation during
anatexis: theoretical constraints, a numerical model, and its application to pelitic
migmatites of the Tickalra Metamorphics, northwestern Australia. Geochimica et
Cosmochimica Acta, 65, 2771–88.
Nesse, W. D. (1986). Introduction to Optical Mineralogy. New York: Oxford
University Press.
Nicolas, A. (1992). Kinematics in magmatic rocks with special reference to gabbros.
Journal of Petrology, 33, 891–915.
Nicolas, A. & Ildefonse, B. (1996). Flow mechanism and viscosity in basaltic magma
chambers. Geophysical Research Letters, 23, 2013–16.
Orford, J. D. & Whalley, W. B. (1983). The use of the fractal dimension to quantify the
morphology of irregular-shaped particles. Sedimentology, 30, 655–68.
Ortoleva, P. J. (1994). Geochemical Self-Organisation. New York: Oxford University
Press.
Pagel, M. (2000). Cathodoluminescence in Geosciences. Berlin: Springer.
Palmer, H. C. & MacDonald, W. D. (1999). Anisotropy of magnetic susceptibility in
relation to source vents of ignimbrites: empirical observations. Tectonophysics,
307, 207–18.
Palmer, H. C. & MacDonald, W. D. (2002). The Northeast Nevada Volcanic Field:
Magnetic properties and source implications. Journal of Geophysical ResearchSolid Earth, 107, (B11), article number 2298.
Pan, Y. (2001). Inherited correlation in crystal size distribution. Geology, 29, 227–30.
Pareschi, M., Pompilio, M. & Innocenti, F. (1990). Automated evaluation of spatial
grain size distribution density from thin section images. Computers &
Geosciences, 16, 1067–84.
Park, H. H. & Yoon, D. N. (1985). Effect of dihedral angle on the morphology of
grains in a matrix phase. Metallurgical Transactions A-Physical Metallurgy and
Materials Science, 16, 923–8.
Parsons, I., ed., (1987). Origins of Igneous Layering. NATO ASI series. Series C,
Mathematical and Physical Sciences. Vol. 196. Dordrecht: D. Reidel Pub. Co.,
Boston/Norwell, MA, USA: Kluwer Academic Publishers.
Pearce, T. H. & Clark, A. H. (1989). Nomarski interference contrast observations of
textural details in volcanic rocks. Geology, 17, 757–9.
Pearce, T. H., Russell, J. K. & Wolfson, I. (1987). Laser-interference and Nomarski
interference imaging of zoning profiles in plagioclase phenocrysts from the May 18,
1980,eruption of Mount St. Helens, Washington. American Mineralogist, 72, 1131–43.
Perring, C. S., Barnes, S. J., Verrall, M. & Hill, R. E. T. (2004). Using automated
digital image analysis to provide quantitative petrographic data on olivine-phyric
basalts. Computers & Geosciences, 30, 183–95.
Peterson, T. D. (1990). Petrology and genesis of natrocarbonatite. Contributions to
Mineralogy and Petrology, 105, 143–55.
Peterson, T. D. (1996). A refined technique for measuring crystal size distributions in
thin section. Contributions to Mineralogy and Petrology, 124, 395–405.
Petford, N., Davidson, G. & Miller, J. A. (2001). Investigation of the petrophysical
properties of a porous sandstone sample using confocal scanning laser
microscopy. Petroleum Geoscience, 7, 99–105.
Petrik, I., Nabelek, P. I., Janak, M. & Plasienka, D. (2003). Conditions of formation
and crystallization kinetics of highly oxidized pseudo tachylytes from the high
Tatras (Slovakia). Journal of Petrology, 44, 901–27.
References
253
Petruk, W. (1989). Image Analysis in Earth Sciences. Ottawa: Mineralogical
Association of Canada.
Philpotts, A. R. & Dickson, L. D. (2000). The formation of plagioclase chains during
convective transfer in basaltic magma. Nature, 406, 59–61.
Philpotts, A. R. & Dickson, L. D. (2002). Millimeter-scale modal layering and the
nature of the upper solidification zone in thick flood-basalt flows and other sheets
of magma. Journal of Structural Geology, 24, 1171–7.
Philpotts, A. R., Shi, J. Y. & Brustman, C. (1998). Role of plagioclase crystal chains in
the differentiation of partly crystallized basaltic magma. Nature, 395, 343–6.
Philpotts, A. R., Brustman, C. M., Shi, J. Y., Carlson, W. D. & Denison, C. (1999).
Plagioclase-chain networks in slowly cooled basaltic magma. American
Mineralogist, 84, 1819–29.
Pickering, G., Bull, J. M. & Sanderson, D. J. (1995). Sampling power-law
distributions. Tectonophysics, 248, 1–20.
Pirard, E. (2004). Multispectral imaging of ore minerals in optical microscopy.
Mineralogical Magazine, 68, 323–33.
Polacci, M., Cashman, K. V. & Kauahikaua, J. P. (1999). Textural characterization
of the pahoehoe-a’a transition in Hawaiian basalt. Bulletin of Volcanology,
60, 595–609.
Poland, M. P., Fink, J. H. & Tauxe, L. (2004). Patterns of magma flow in segmented
silicic dikes at Summer Coon volcano, Colorado: AMS and thin section analysis.
Earth and Planetary Science Letters, 219, 155–69.
Prince, C. M., Ehrlich, R. & Anguy, Y. (1995). Analysis of spatial order in sandstones.
2. Grain clusters, packing flaws, and the small-scale structure of sandstones.
Journal of Sedimentary Research Section A-Sedimentary Petrology and Processes,
65, 13–28.
Prior, D. J., Boyle, A. P., Brenker, F. et al. (1999). The application of electron
backscatter diffraction and orientation contrast imaging in the SEM to textural
problems in rocks. American Mineralogist, 84, 1741–59.
Prior, D. J., Wheeler, J., Peruzzo, L., Spiess, R. & Storey, C. (2002). Some garnet
microstructures: an illustration of the potential of orientation maps and
misorientation analysis in microstructural studies. Journal of Structural Geology,
24, 999–1011.
Proussevitch, A. A. & Sahagian, D. L. (2001). Recognition and separation of discrete
objects within complex 3D voxelized structures. Three-Dimensional
Reconstruction, Modelling and Visualization of Geological Materials, 27, 441–54.
Pupin, J. P. (1980). Zircon and granite petrology. Contributions to Mineralogy and
Petrology, 73, 207–20.
Randle, V. & Caul, M. (1996). Representation of electron backscatter diffraction data.
Materials Science and Technology, 12, 844–50.
Randle, V. & Engler, O. (2000). Introduction to Texture Analysis: Macrotexture,
Microtexture and Orientation Mapping. Reading, UK: Gordon and Breach
Science Publishers.
Randolph, A. D. & Larson, M. A. (1971). Theory of Particulate Processes. New York:
Academic Press.
Reed, S. J. B. (1996). Electron Microprobe Analysis and Scanning Electron Microscopy
in Geology. Cambridge, New York: Cambridge University Press.
Reimann, C. & Filzmoser, P. (2000). Normal and lognormal data distribution in
geochemistry; death of a myth; consequences for the statistical treatment of
geochemical and environmental data. Environmental Geology, 39, 1001–14.
254
References
Resmini, R. G. & Marsh, B. D. (1995). Steady-state volcanism, paleoeffusion rates,
and magma system volume inferred from plagioclase crystal size distributions in
mafic lavas; Dome Mountain, Nevada. Journal of Volcanology and Geothermal
Research, 68, 273–96.
Riegger, O. & van Vlack, L. (1960). Dihedral angle measurement. Metallurgical
Society of the American Institute of Metallurgical Engineers Transactions, 218,
933–5.
Robin, P. Y. F. (2002). Determination of fabric and strain ellipsoids from measured
sectional ellipses – theory. Journal of Structural Geology, 24, 531–44.
Rochette, P., Aubourg, C. & Perrin, M. (1999). Is this magnetic fabric normal?
A review and case studies in volcanic formations. Tectonophysics, 307, 219–34.
Rochette, P., Jackson, M. & Aubourg, C. (1992). Rock magnetism and the
interpretation of anisotropy of magnetic-susceptibility. Reviews of Geophysics,
30, 209–26.
Rodriguez-Navarro, A. B. & Romanek, C. S. (2002). Mineral fabrics analysis using a
low-cost universal stage for X-ray diffractometry. European Journal of
Mineralogy, 14, 987–92.
Rogers, C. D. F., Dijkstra, T. A. & Smalley, I. J. (1994). Particle packing from an
earth-science viewpoint. Earth-Science Reviews, 36, 59–82.
Rombouts, L. (1995). Sampling and statistical evaluation of diamond deposits.
Journal of Geochemical Exploration, 53, 351–67.
Rosenberg, C. L. & Handy, M. R. (2005). Experimental deformation of partially
melted granite revisited: implications for the continental crust. Journal of
Metamorphic Geology, 23, 19–28.
Ross, B. J., Fueten, F. & Yashkir, D. Y. (2001). Automatic mineral identification
using genetic programming. Machine Vision and Applications, 13, 61–9.
Royet, J.-P. (1991). Stereology: A method for analysing images. Progress in
Neurobiology, 37, 433–74.
Rubin, A. E. (2000). Petrologic, geochemical and experimental constraints on models
of chondrule formation. Earth-Science Reviews, 50, 3–27.
Rubin, A. E. & Grossman, J. N. (1987). Size-frequency-distributions of EH3
chondrules. Meteoritics, 22, 237–51.
Rudashevsky, N. S., Burakov, B. E., Lupal, S. D., Thalhammer, O. A. R. &
Sainieidukat, B. (1995). Liberation of accessory minerals from various rock types
by electric-pulse disintegration-method and application. Transactions of the
Institution of Mining and Metallurgy Section C-Mineral Processing and Extractive
Metallurgy, 104, C25–C29.
Russ, J. C. (1986). Practical Stereology. New York: Plenum Press.
Russ, J. C. (1999). The Image Processing Handbook. Boca Raton, Florida, USA: CRC
Press.
Rust, A. C. & Cashman, K. V. (2004). Permeability of vesicular silicic magma: inertial
and hysteresis effects. Earth and Planetary Science Letters, 228, 93–107.
Rust, A. C., Manga, M. & Cashman, K. V. (2003). Determining flow type, shear rate
and shear stress in magmas from bubble shapes and orientations. Journal of
Volcanology and Geothermal Research, 122, 111–32.
Sahagian, D. L. & Maus, J. E. (1994). Basalt vesicularity as a measure of atmosphericpressure and palaeoelevation. Nature, 372, 449–51.
Sahagian, D. L. & Proussevitch, A. A. (1998). 3D particle size distributions from 2D
observations; stereology for natural applications. Journal of Volcanology and
Geothermal Research, 84, 173–96.
References
255
Sahagian, D. L., Proussevitch, A. A. & Carlson, W. D. (2002a). Analysis of vesicular
basalts and lava emplacement processes for application as a paleobarometer/
paleoaltimeter. Journal of Geology, 110, 671–85.
Sahagian, D. L., Proussevitch, A. A. & Carlson, W. L. (2002b). Timing of Colorado
Plateau uplift: Initial constraints from vesicular basalt-derived paleoelevations.
Geology, 30, 807–10.
Saiki, K. (1997). Morphology and simulation of solid state rounding process.
Geophysical Research Letters, 24, 1519–22.
Saiki, K., Laporte, D., Vielzeuf, D., Nakashima, S. & Boivin, P. (2003).
Morphological analysis of olivine grains annealed in an iron-nickel matrix:
Experimental constraints on the origin of pallasites and on the thermal history of
their parent bodies. Meteoritics & Planetary Science, 38, 427–44.
Saint-Blanquat, M. & Tikoff, B. (1997). Development of magmatic to solid-state
fabrics during syntectonic emplacement of the Mono Creek granite, Sierra
Nevada Batholith. In J. L. Bouchez, D. H. W. Hutton, & W. E. Stephens, eds.,
Granite: From Segregation of Melt to Emplacement Fabrics. Dordrecht: Kluwer
Academic Publishers, pp. 231–52.
Saltikov, S. A. (1967). The determination of the size distribution of particles in an
opaque material from a measurement of the size distributions of their sections. In
H. Elias, ed., Proceedings of the Second International Congress for Stereology.
Berlin: Springer-Verlag, pp. 163–73.
Saltzer, R. L., Chatterjee, N. & Grove, T. L. (2001). The spatial distribution of garnets
and pyroxenes in mantle peridotites: Pressure-temperature history of peridotites
from the Kaapvaal craton. Journal of Petrology, 42, 2215–29.
Sato, H. (1995). Textural difference between pahoehoe and a’a lavas of Izu-Oshima
volcano, Japan; an experimental study on population density of plagioclase.
Models of Magmatic Processes and Volcanic Eruptions, 66, 101–13.
Schafer, F. & Foley, S. F. (2002). The effect of crystal orientation on the wetting
behaviour of silicate melts on the surfaces of spinel peridotite minerals.
Contributions to Mineralogy and Petrology, 143, 254–61.
Schafer, W. (2002). Neutron diffraction applied to geological texture and stress
analysis. European Journal of Mineralogy, 14, 263–89.
Schmid, S. M., Casey, M. & Starkey, J. (1981). An illustration of the advantages of a
complete texture analysis described by the orientation distribution function
(ODF) using quartz pole figure data. Tectonophysics, 78, 101–17.
Schwindinger, K. R. (1999). Particle dynamics and aggregation of crystals in a magma
chamber with application to Kilauea Iki olivines. Journal of Volcanology and
Geothermal Research, 88, 209–38.
Schwindinger, K. R. & Anderson, A. T. (1989). Synneusis of Kilauea Iki olivines.
Contributions to Mineralogy and Petrology, 103, 187–98.
Scott, R. G. & Benn, K. (2001). Peak-ring rim collapse accommodated by impact
melt-filled transfer faults, Sudbury impact structure, Canada. Geology, 29,
747–50.
Sempels, J.-M. (1978). Evidence for constant habit development of plagioclase
crystals from igneous rocks. Canadian Mineralogist, 16, 257–63.
Shelley, D. (1985). Determining paleo-flow directions from groundmass fabrics in the
Lyttelton radial dykes, New Zealand. Journal of Volcanology and Geothermal
Research, 25, 69–79.
Shin, H., Lindquist, W. B., Sahagian, D. L. & Song, S. R. (2005). Analysis of the
vesicular structure of basalts. Computers & Geosciences, 31, 473–87.
256
References
Shore, M. & Fowler, A. D. (1999). The origin of spinifex texture in komatiites. Nature,
397, 691–4.
Smit, T. H., Schneider, E. & Odgaard, A. (1998). Star length distribution: a volumebased concept for the characterization of structural anisotropy. Journal of
Microscopy-Oxford, 191, 249–57.
Smith, C. S. (1948). Grains, phases and interfaces: an interpretation of microstructure.
Transactions of the Metallurgical Society of the AIME, 175, 15–51.
Smith, C. S. (1964). Some elementary principles of polycrystalline microstructure.
Metallurgical Review, 9, 1–48.
Smith, J. V. (2002). Structural analysis of flow-related textures in lavas. Earth-Science
Reviews, 57, 279–97.
Song, S. R., Jones, K. W., Lindquist, W. B., Dowd, B. A. & Sahagian, D. L. (2001).
Synchrotron X-ray computed microtomography: studies on vesiculated basaltic
rocks. Bulletin of Volcanology, 63, 252–63.
Sparks, R. S. J. (1978). Dynamics of bubble formation and growth in magmas – review
and analysis. Journal of Volcanology and Geothermal Research, 3, 1–37.
Sparks, R. S. J. (1997). Causes and consequences of pressurisation in lava dome
eruptions. Earth and Planetary Science Letters, 150, 177–89.
Spiess, R., Peruzzo, L., Prior, D. J. & Wheeler, J. (2001). Development of garnet
porphyroblasts by multiple nucleation, coalescence and boundary
misorientation-driven rotations. Journal of Metamorphic Geology, 19, 269–90.
Stamatelopoulou-Seymour, K., Vlassopoulos, D., Pearce, T. H. & Rice, C. (1990).
The record of magma chamber processes in plagioclase phenocrysts at Thera
volcano, Aegean Volcanic Arc, Greece. Contributions to Mineralogy and
Petrology, 104, 73–84.
Stipp, M. & Tullis, J. (2003). The recrystallized grain size piezometer for quartz.
Geophysical Research Letters, 30, article number 2088.
Sturm, R. (2004). Imaging of growth banding of minerals using 2-stage sectioning:
application to accessory zircon. Micron, 35, 681–4.
Sunagawa, I. (1987a). Morphology of Crystals. Dordrecht: D. Reidel.
Sunagawa, I. (1987b). Morphology of minerals. In I. Sunagawa, ed., Morphology of
Crystals. Dordrecht: D. Reidel, pp. 509–88.
Suteanu, C. & Kruhl, J. H. (2002). Investigation of heterogeneous scaling intervals
exemplified by sutured quartz grain boundaries. Fractals-Complex Geometry
Patterns and Scaling in Nature and Society, 10, 435–49.
Suteanu, C., Zugravescu, D. & Munteanu, F. (2000). Fractal approach of structuring
by fragmentation. Pure and Applied Geophysics, 157, 539–57.
Sutton, A. P. & Balluffi, R. W. (1996). Interfaces in Crystalline Materials. Oxford,
UK: Oxford Science Publications.
Swan, A. R. H. & Sandilands, M. (1995). Introduction to Geological Data Analysis.
Cambridge, MA, USA: Blackwell Science.
Syvitski, J. P. M. (1991). Principles, Methods, and Application of Particle Size Analysis.
New York: Cambridge University Press.
Syvitski, J. P. M., Asprey, K. & Clattenberg, D. (1991). Principles, design, and calibration
of settling tubes. In J. P. M. Syvitski, ed., Principles, Methods, and Application of
Particle Size Analysis. New York: Cambridge University Press, pp. 45–63.
Takahashi, M. & Nagahama, H. (2000). Fractal grain boundary migration. Fractals,
8, 189–94.
Tarling, D. H. & Hrouda, F. (1993). The Magnetic Anisotropy of Rocks. London:
Chapman & Hall.
References
257
Tarquini, S. & Armienti, P. (2001). Film color scanner as a new and cheap tool
for image analysis in petrology. Image Analysis and Stereology, 20 (Suppl. 1),
567–72.
Taylor, L. (2000). Diamonds and their mineral inclusions, and what they tell us: A
detailed pull-part of a diamondiferous eclogite. International Geology Review, 42,
959–83.
Taylor, L. A., Nazarov, M. A., Shearer, C. K. et al. (2002). Martian meteorite Dhofar
019: A new shergottite. Meteoritics & Planetary Science, 37, 1107–28.
Thomas, M. C., Wiltshire, R. J. & Williams, A. T. (1995). The use of Fourier
descriptors in the classification of particle-shape. Sedimentology, 42, 635–45.
Thomas-Keprta, K. L., Bazylinski, D. A., Kirschvink, J. L. et al. (2000). Elongated
prismatic magnetite crystals in ALH84001 carbonate globules: Potential Martian
magnetofossils. Geochimica et Cosmochimica Acta, 64, 4049–81.
Thompson, S., Fueten, F. & Bockus, D. (2001). Mineral identification using artificial
neural networks and the rotating polarizer stage. Computers & Geosciences, 27,
1081–9.
Titkov, S. V., Saparin, G. V. & Obyden, S. K. (2002). Evolution of growth sectors in
natural diamond crystals as revealed by cathodoluminescence topography.
Geology of Ore Deposits, 44, 350–60.
Toramaru, A. (1989). Vesiculation process and bubble size distributions in ascending
magmas with constant velocities. Journal of Geophysical Research, B, Solid Earth
and Planets, 94, 17523–42.
Toramaru, A. (1990). Measurement of bubble-size distributions in vesiculated rocks
with implications for quantitative estimation of eruption processes. Journal of
Volcanology and Geothermal Research, 43, 71–90.
Treiman, A. H. (2003). Submicron magnetite grains and carbon compounds in
Martian meteorite ALH84001: Inorganic, abiotic formation by shock and
thermal metamorphism. Astrobiology, 3, 369–92.
Trindade, R. I. F., Bouchez, J. L., Bolle, O. et al. (2001). Secondary fabrics revealed by
remanence anisotropy: methodological study and examples from plutonic rocks.
Geophysical Journal International, 147, 310–18.
Tuffen, H. (1998). L’origine des cristaux dans le chambre magmatique de Santorin
(Grèce). Clermont-Ferrand, France: Université Blaise-Pascal.
Turcotte, D. L. (1992). Fractals and Chaos in Geology and Geophysics. Cambridge,
New York: Cambridge University Press.
Turner, S., George, R., Jerram, D. A., Carpenter, N. & Hawkesworth, C. (2003). Case
studies of plagioclase growth and residence times in island arc lavas from Tonga
and the Lesser Antilles, and a model to reconcile discordant age information.
Earth and Planetary Science Letters, 214, 279–94.
Underwood, E. E. (1970). Quantitative Stereology. Reading, MA: Addison-Wesley.
van den Berg, E. H., Meesters, A. G. C. A., Kenter, J. A. M. & Schlager, W. (2002).
Automated separation of touching grains in digital images of thin sections*1.
Computers & Geosciences, 28, 179–90.
Vance, J. A. (1969). On synneusis. Contributions to Mineralogy and Petrology, 24,
7–29.
Vavra, G. (1993). A guide to quantitative morphology of accessory zircon. Chemical
Geology, 110, 15–28.
Ventura, G. (2001). The strain path and emplacement mechanism of lava flows: an
example from Salina (southern Tyrrhenian Sea, Italy). Earth and Planetary
Science Letters, 188, 229–40.
258
References
Ventura, G., DeRosa, R., Colletta, E. & Mazzuoli, R. (1996). Deformation patterns in
a high-viscosity lava flow inferred from the crystal preferred orientation and
imbrication structures: An example from Salina (Aeolian Islands, southern
Tyrrhenian Sea, Italy). Bulletin of Volcanology, 57, 555–62.
Vernon, R. (1970). Comparative grain boundary studies of some basic and
ultrabasic granulites, nodules and cumulates. Scottish Journal of Geology,
6, 337–51.
Vernon, R. H. (1968). Microstructures of high-grade metamorphic rocks at Broken
Hill, Australia. Journal of Petrology, 9, 1–22.
Vernon, R. H. (1986). K-feldspar megacrysts in granites – phenocrysts not
porphyroblasts. Earth-Science Reviews, 23, 1–63.
Vernon, R. H. (2004). A Practical Guide to Rock Microstructure. Cambridge:
Cambridge University Press.
Verrecchia, E. P. (2003). Foreword: image analysis and morphometry of geological
objects. Mathematical Geology, 35, 759–62.
Vigneresse, J. L., Barbey, P. & Cuney, M. (1996). Rheological transitions during
partial melting and crystallization with application to felsic magma segregation
and transfer. Journal of Petrology, 37, 1579–1600.
Voorhees, P. W. (1992). Ostwald ripening of two-phase mixtures. Annual Review of
Materials Science, 22, 197–215.
Wada, Y. (1992). Magma flow directions inferred from preferred orientations of
phenocryst in a composite feeder dyke, Miyake-Jima, Japan. Journal of
Volcanology and Geothermal Research, 49, 119–26.
Waff, H. S. & Bulau, J. R. (1979). Equilibrium fluid distribution in an ultramafic
partial melt under hydrostatic stress conditions. Journal of Geophysical Research,
84, 6109–14.
Wager, L. R. (1961). A note on the origin of ophitic texture in the chilled olivine
gabbro of the Skaergaard intrusion. Geological Magazine, 98, 353–66.
Wager, L. R. & Brown, G. M. (1968). Layered Igneous Rocks. Edinburgh; London:
Oliver & Boyd.
Waters, C. & Boudreau, A. E. (1996). A re-evaluation of crystal size distribution in
chromite cumulates. American Mineralogist, 81, 1452–9.
Watson, E. B. & Brenan, J. M. (1987). Fluids in the lithosphere.1. experimentallydetermined wetting characteristics of CO2-H2O fluids and their implications for
fluid transport, host-rock physical-properties, and fluid inclusion formation.
Earth and Planetary Science Letters, 85, 497–515.
Wegner, M. & Christie, J. (1985). General chemical etchants for microstructures and
defects in silicates. Physics and Chemistry of Minerals, 12, 90–2.
Wenk, H. R. (2002). Texture and anisotropy. In Plastic Deformation of Minerals and
Rocks. Reviews in Mineralogy & Geochemistry, 51. Washington DC:
Mineralogical Society of America, pp. 291–329.
Wenk, H. R. (1985). Preferred Orientation in Deformed Metals and Rocks: An
Introduction to Modern Texture Analysis. Orlando FL: Academic Press.
Wenk, H. R. & Grigull, S. (2003). Synchrotron texture analysis with area detectors.
Journal of Applied Crystallography, 36, 1040–9.
Wenk, H. R. & Van Houtte, P. (2004). Texture and anisotropy. Reports on Progress in
Physics, 67, 1367–428.
Whitham, A. & Sparks, R. (1986). Pumice. Bulletin of Volcanology, 48, 209–23.
Wilson, B., Dewers, T., Ze’ev, R. & Brune, J. (2005). Particle size and energetics of
gouge from earthquake rupture zones. Nature, 434, 749–52.
References
259
Wright, I. C., Gamble, J. A. & Shane, P. A. R. (2003). Submarine silicic volcanism of
the Healy caldera, southern Kermadec arc (SW Pacific): I – volcanology and
eruption mechanisms. Bulletin of Volcanology, 65, 15–29.
Wulff, G. (1901). Zur frage der Geschwindigkeit des Wachstums und der Auflosumg
der Krystallflachen. Zeitschift für Kristallographie und Mineralogie, 34, 449–530.
Xie, Y. X., Wenk, H. R. & Matthies, S. (2003). Plagioclase preferred orientation by
TOF neutron diffraction and SEM-EBSD. Tectonophysics, 370, 269–86.
Yaouancq, G. & MacLeod, C. J. (2000). Petrofabric investigation of gabbros from the
Oman ophiolite: comparison between AMS and rock fabric. Marine Geophysical
Researches, 21, 289–305.
Zeh, A. (2004). Crystal size distribution (CSD) and textural evolution of accessory
apatite, titanite and allanite during four stages of metamorphism: an example
from the Moine supergroup, Scotland. Journal of Petrology, 45, 2101–32.
Zellmer, G., Turner, S. & Hawkesworth, C. (2000). Timescales of destructive plate
margin magmatism; new insights from Santorini, Aegean volcanic arc. Earth and
Planetary Science Letters, 174, 265–81.
Zhou, Y., Starkey, J. & Mansinha, L. (2004a). Identification of mineral grains in a
petrographic thin section using phi- and max-images. Mathematical Geology,
36, 781–801.
Zhou, Y., Starkey, J. & Mansinha, L. (2004b). Segmentation of petrographic images
by integrating edge detection and region growing. Computers & Geosciences,
30, 817–31.
Zieg, M. J. & Marsh, B. D. (2002). Crystal size distributions and scaling laws in the
quantification of igneous textures. Journal of Petrology, 43, 85–101.
Index
a’a 106, 186
accessory minerals 130
adcumulus growth 66
allanite 130, 159
andesite 108–11, 185
anisotropy of anhysteretic remanent
magnetisation 179–80
anisotropy of magnetic susceptibility (AMS) 177–9,
187–8, 192–4
annealing 60
anorthoclase 113
anorthosite 116–18, 156
apatite 130, 159
artefacts 16–17
ASTM mean size 77–8
autocorrelation function 174
automatic image analysis 28–32
AVA diagrams 182
Avrami equations 48
backscattered electron images 18
basalt 105–7, 162, 187
Benford plate 23, 32–3
bimrocks 133
Bishop Tuff 121, 231
broken crystals and blocks 15, 124, 131–4
bubbles 219–21, 228–31
Bushveld complex 193, 213
calcite 132, 161
carbonate 128
carbonatite 126
cataclasis 70–1
Cathedral Peak granodiorite 123–4
cathodoluminescence 20
characteristic length 50
chi-squared test 36
chondrule 127
chromite 126
circularity 153
classified image 28, 33
closure in crystal size distribution 71
cluster analysis 207–8
coalescence 130, 166, 197, 220
coarsening 59–68
accessory minerals 130
adcumulus growth 66
bubbles 221
chromite 126
Communicating Neighbours model 64–5, 116
dihedral angles 142
driving force 59–60
energy considerations 60–2
equilibrium concentrations 61
experimental study 129
fluid focusing 65–6
growth rate dispersion 65
interfacial free energy 60
lava lake 105
Lifshitz–Slyozov–Wagner model 62–4, 116
mafic plutonic rocks 115
material transport 62
metamorphic rocks 68, 130
open systems 65–6
orientation 170, 193
second-phase effects 68
silicic volcanic rocks 122
slabs 25–6
solid materials 68
surface energy 60
volcanic rocks 111–14
Zener-influenced coarsening 68
coercivity 180
colonnade 106
comb-layering 166
Communicating Neighbours coarsening model
64–5, 116
compaction 52–4
mechanical compaction 52–3
pressure solution compaction 53–4, 170,
192
competitive particle growth 60
computed axial tomography (CAT) 12
computer-integrated polarisation (CIP)
microscopy 32–3
extinction angle image 33
260
Index
maximum birefringence image 33
orientation 175–6
confocal microscopy 12
connectivity of melts 142
contiguity of phases 34
cordierite 131
crystal accumulation and fractionation
55–6, 59
crystal chains 193, 209, 214
crystal deformation and destruction
deformation 68–70
fluid inclusions 70
fractals 70–1
fragmentation 70–1
fragmentation models 71
grain size reduction 68
lattice defects 69
mosaic texture 70
recovery 70
strain 70
crystal growth 46–8
advection 48
driving force 46
growth rate 43, 48
mean projected height 42
mechanism 47
roughening transition 47
saturation 46
transport and supply 48
undercooling 46
crystal habit 135
crystal size distribution 40
bi-logarithmic 98
closure 71
coarsening 59–68
crystal nucleation 44–6
deformed and broken crystals
deformation 68–70
empty bins 97
equilibration 59–68
graphs 95–101
histograms 97
intersection data, extraction of size
distribution 79–91
kinetic textures 48–52
kinetic versus equilibrium effects 42–3
lognormal 99–101
mafic plutonic rocks 114–19
mafic volcanic rocks 103–14
mechanically modified textures 52–9
moments of number density 102
nucleation and growth rates 43
outline definitions 76–7
primary kinetic processes 42–8
projected data, extraction of size distribution 92
semi-logarithmic diagram 96–8
silicic plutonic rocks 122–4
silicic volcanic rocks 119
size distribution theory 42–3, 74
three-dimensional analytical methods 74–5
two-dimensional analytical methods 76–7
undeformed metamorphic minerals 129–31
unimodal and bimodal 40
cumulative distribution function 182, 184
dacite 119–20, 185, 189
Darcy’s law 217
data brick 13
data quality 3
data volumes 13–14
seeded threshold filter 14
touching crystals 14
watershed algorithm 14
deformation 68–70, 168
destructive analytical methods 14–16
disaggregation 15
dissolution 15–16
electric pulse disintegration 15
diamonds 15, 124–5
differential interference contrast (DIC)
microscopy 24–5
dihedral angle 44, 139–43
applications 161–4
impingement 140
measurement 154
pores 218–19
directional variogram 174
disaggregation 15
Dissector method 78
dissipative structures 201
dissolution 15–16
Dome Mountain 108
dykes 188–90
dynamic recrystallisation 132
eclogite 125
edge effects 9, 223
electric pulse disintegration 15
electron microprobe 17
electronic analytical methods 17–20
backscattered electron images 18
cathodoluminescence 20
orientation contrast imaging 20
X-ray maps 19
enclaves 170, 185
engineering 132, 196
entablature 106
equal-area net 181
equilibration 59–68
etching 25
exobiology 128
experimental studies 129
explosions 132
external surface area 216
extinction angle image 33
fabric, definition 3
fault gouge 134
Feret diameter 77
Feret length 77
filter-pressing 52–4
fines destruction 60
261
262
flotation 113
fluid focusing 65–6
fluid inclusions 70
fluid, definition 48
fossil bacteria 128
Fourier analysis 151–3, 182, 210
Fourier descriptors 152
fractals 70–1, 131, 150–1, 160, 228
fractured crystals 15, 124, 131–4
fragmentation 70–1, 132–4
frameworks 212
gabbro 114–16, 156
garnet 129–30, 178, 196
Gaussian diffusion-resample method 151
genetic programming 32
geochemical self-organisation 46, 201–2
global parameters 33–5
contiguity 34
goodness-of-fit 37
intercept method 35
interface density 34, 79
neighbourhood 35
volume fraction 34
grain size distribution (sediments) 40
analytical methods 75
electrolyte 75
laser methods 75
parametric solutions 81
settling rate 75
sieves 75
grain size limits 9
grain size reduction 68
grain-boundary area reduction 60, 68
grain-boundary migration 48, 60, 139
granite 162, 215
granodiorite 123–4
granulite 161, 212
Grey data conversion method 91–2
grey-scale images 28–30
growth rate 107–8, 129
growth-rate dispersion 65
habit 135
Halle volcanic complex 122, 195, 213
Hawaii 104, 112–14, 230
Haywood circular parameter 149
heavy liquids 16
helium pycnometry 224
hematite 128
hue-saturation-intensity model 30
hydrothermal crystals 131
ilmenite 192
image analysis 26–33
automatic image analysis 28–32
classified image measurement 33
edge effects 9
grey-scale images 28–30
image classification 28
manual image analysis 27–8
Index
multichannel images 30–2
noise 32
size limits of measurements 27
touching grains 32
training 29, 31
imbrication 169, 172, 188
impact melt rocks 118–19
impingement 140
information density 4
infrared adsorption 23
intercept method 35
interface density 34
interfacial energy or tension, see surface energy
internal surface area 216
intersection data, extraction of size distribution
79–91
intersection size data conversion 79–91
Inyo Domes 120
Kameni islands 119
Kiglapait intrusion 114, 191
kimberlite 133
komatiite 214
LA (Los Angeles) index 196
Lac-Saint-Jean anorthosite suite 116
lamination 117
lamproite 125
lattice defects 69, 138–9
lattice-preferred orientation methods 174–7
computer-integrated polarisation microscopy
175–6
electron backscatter diffraction 176–7
optical analytical methods 174–5
lava flows 103–14, 119
lava lake 103
layering 116, 191
Lebel syenite 195
lherzolite 131
Liesegang banding 202
Lifshitz–Slyozov–Wagner coarsening model 62–4,
116, 130
linear attenuation coefficient 13
literature sources 4
lognormal size distribution 99–101, 121,
127, 131
magma deformation 156
magma mixing 56, 119, 204
magnetic orientation methods
anisotropy of anhysteretic remanent
magnetisation 179–80
anisotropy of magnetic susceptibility 177–9,
187–8
natural remanent magnetism 179
partial anisotropy of anhysteretic remanent
magnetisation 179–80, 192
magnetic resonance imaging 13
magnetic susceptibility 177
magnetite 128, 178–9
Mandelbrot–Richardson diagram 150
Index
mantle 131, 215
manual image analysis 27–8
Mars 128
maximum birefringence image 33
maximum likelihood classifier 32
McKenzie dyke swarm 189
mean atomic number 18
megacrysts 123–4
mercury intrusion porosimetry 223
meteorite 127–8, 159
microcline 123–4
microlites 119–20, 155, 188
microstructure 3
micro-XRF 19
migmatite 131
mineral point correlations 210–11
minimum energy melt fraction 143
moments 102
Mono Creek granitoid 195
morphology 135
Mt Erebus 113
Mt Etna 105
Mt Mazama 229
Mt Pinatubo 121
Mt St Helens 119
Mt Taranaki 108
multichannel images 30–2
genetic programming 32
hue-saturation-intensity model 30
maximum likelihood classifier 32
narrow bandwidth filters 24, 30
neural networks 32
principal component analysis 30–1
red–blue–green model 30
mylonites 161
natural remanent magnetism 179
nearest-neighbour distance (big R) 206–7
neural networks 32
neutron diffraction 180
nitrogen adsorption method 223
Nomarski (DIC) microscopy 24–5
nucleation, crystals and bubbles 44–6, 219–21
critical radius 44
dihedral angle 44
heterogeneous 44, 219
homogeneous 44, 219
position 200–1
spinoidal decomposition 45
undercooling 43
null hypothesis 36
oikocrysts 110, 116
Oldoinyo Lengai 126
olivine
dihedral angles 161–2
orientation 192
shape 158–9
size 112–14, 131–2
Oman ophiolite 114, 193
ophiolite 114, 190, 193
optical analytical methods 21–5
reflected light microscopy 23–4
transmitted light microscopy 22–3
optical scanning 11
orientation
coarsening 170
dyke flow 188–90
ellipsoid 181
engineering studies 196
felsic lava flows 188
flow in lavas 184–8
granitoids 194–5
individual grain orientation display 181–2
integration of multiple sections 183–4
lattice-preferred orientation
methods 174–7
mafic lava flows 185–8
mafic plutonic rocks 191–4
magmatic crystallisation 166
magmatic deformation 168–70
mean orientations 182–3
plutonic rocks 191–6
pressure-solution compaction 170
pyroclastic rocks 197–8
shape-preferred orientation methods 171–4
solid-state crystallisation 167
solid-state deformation 196–7
solidification 167–8
three-dimensional bulk methods 177–81
orientation contrast imaging 20
orthoclase 122, 213
Ostwald ripening 60
outcrops 26
outline definitions 76–7
packing 204
pahoehoe 106, 186
palaeoelevation of an eruption 230
palaeomagnetism 179
pallasite 159
partial anisotropy of anhysteretic remanent
magnetisation 179–80, 192
percolation threshold 218
peridotite 131, 215
periodic bond chains 136
permeability 66, 217
petrological methods 1–2
phonolite 113
picrite 112–14
plagioclase
crystal chains 193, 209, 214
dihedral angle 163
growth rate 107–8
orientation 185–6, 191
position 213
shape 117, 155–7
size, metamorphic rocks 131
size, plutonic rocks 114–16, 118–19
size, volcanic rocks 103–11, 119–20,
122, 156
polycrystal 3
263
264
pore textures
applications 228–31
bubble shape 222
bubble size distribution 221–2
bubbles, nucleation and growth 219–21
connectedness of pores 227
conversion of 2-D data 225–6
definition of pores 216, 222
display of pore size distributions 226–7
image analysis methods 223
magmas 217–18
moments of pore size distribution 227–8
partially molten rocks 217–18
permeability 217, 226
shapes 228
surface area measurements 223–4
three-dimensional methods 224–5
total porosity 224
two-dimensional methods 225
porphyry 122
position
aggregation and agglutination 203–4
box-counting method 211
cluster analysis 207–8
definitions 204
geochemical self-organisation 201–2
igneous rocks 212–15
magma mixing 204
metamorphic rocks 212
mineral point correlations 210–11
nearest-neighbour distance (big R) 206–7
nucleation of crystals 200–1
packing 204
section and surface methods 205
three-dimensional methods 205
touching crystals 208–9
wavelet and Fourier methods 210
potassium feldspar 122–4, 131
principal component analysis 30–1
probabilities, interpretation 36
projected data, extraction of size distribution 92
projected outline effect 92
projected size data conversion
pseudotachylyte 128
pyroclastic rocks 197–8, 229
pyroxene
dihedral angles 161, 163
position 212–13
shape 159
size 114, 120
qualitative data 2–3
quartz
dihedral angle 162
position 213, 215
shape 158, 160
size 121–2, 132
rapid growth textures 158
recrystallisation 60, 132
reflectance 25
Index
reflected light microscopy 23–4
narrow bandwidth filters 24
Nomarski (DIC) microscopy 24–5
reflectance 25
residence time 50
resolution 9
reticular density 136
roughening transition 47
roundness 153
Saltikov method 85–91
scanning electron microscope 17
seeded threshold filter 14
Sept Iles intrusive suite 116, 191
serial sectioning 10–11
settling of crystals 55–6, 59, 116
settling of sediments, analysis 75
shape 135–64
axial ratio definitions 145
deformed crystals 138–9
deformed rocks 160–1
dihedral angles 139–43, 154, 161–4
equilibrium crystal forms 136–7
extraction of axial ratios 145–8
extraction of grain boundary parameters 149–53
Fourier analysis 151–3
fractal boundary parameters 149–53
Gaussian diffusion-resample method 151
growth crystal forms 137
intersection data treatment 145–8
projection data treatment 148
sedimentological parameters 153
solution crystal forms 137–8
three-dimensional analytical methods 143–4
two-dimensional analytical methods 144–5
undeformed rocks 154–9
shape-preferred orientation methods 171–4
direct 3-D methods 171
extraction of orientation parameters 172–4
inertia tensor method 172
intercept method 172–3
section and surface methods 171–2
star length and volume 173
wavelet methods 174
shear 168, 231
shergottite 127
sieves 75
size 39–134
analytical methods 74–7
definition, three-dimensional 39–40
distribution 40
kinetic textures 48–52
batch crystallisation 51–2
Gaussian rate variations 52
growth probability model 52
size-proportional growth rate 52
steady-state models 49–51
limits of measurements 9, 27
linear size definition 40–2
mean projected height 42
mean size 40, 77–9
Index
mechanically modified textures 52–9
accumulation and fractionation 55–6, 59, 116
compaction 52–4
magma mixing 56, 119
number density 40
slope and intercept correlation, CSD 73
software 5–6
solidification 167–8
solutions 48
Soufriere Hills, Montserrat 109
sphericity 145
spherulitic crystals 137
spinifex 166
spinoidal decomposition 45
staining 25
steady-state crystallisation models 49–51
changes in parameters 51
characteristic length 50
cooling equations 50
impact melt 118–19
lava lake 103–5
residence time 50
stereology 8, 33, 80
biased or inaccessible parameters 34
correction of data 93–5
cut-section effect 83
discredited method 91
distribution-free solutions 82–5
exact, unbiased or accessible parameters 34
global parameters 33–5
intersection-probability effect 82
parametric solutions 81–2
projected outlines 92
Saltikov method 85–91
stereonet 181
Stillwater complex 114, 126, 202, 215
straight line distributions 37–8
strain 70
Stripstar data conversion method 90
Sudbury impact structure 118–19, 190
sulphide 163
surface area per unit volume 34, 79, 223
surface energy 60, 136, 141–2
surface preparation 16–17
sutured grain boundaries 139, 160
synchrotron radiation 13, 180
synneusis 166, 203
Tellnes ilmenite deposit 192
textural development sequences 9–10
textural equilibrium 142
textural maturation 60
texture 3
Thera island group 119
three-dimensional analytical methods 10–13
confocal microscopy 12
magnetic resonance imaging 13
optical scanning 11
serial sectioning 10–11
X-ray tomography 12
titanite 130
tomography 12
total grain number 78
touching crystals 14, 32
training in image analysis 29, 31
transmitted light microscopy 22–3
Benford plate 23
computer-integrated polarisation (CIP)
microscopy 32–3
infrared adsorption 23
resolution 22
universal stage 23
whole thin sections 22
trapped liquid 218
troctolite 116–18
Troodos ophiolite 190
undercooling 43, 46
universal stage 23, 175
verification of distributions 35–8
chi-squared test 36
goodness-of-fit 37
null hypothesis 36
probabilities, interpretation 36
straight line distributions 37–8
volumetric phase proportion 34, 93–5
voxels 13
Wager data conversion method 91, 103
watershed algorithm 14, 32
wavelet 153, 174, 210
Weibull distribution 127
wetting of surfaces 142
whole thin section imaging 22
xenocrysts 112, 124
X-ray diffraction 180
X-ray fluorescence 19
X-ray maps 19, 30
X-ray tomography 12
linear attenuation coefficient 13
synchrotron radiation 13
zircon 121–2, 157–8
zoning patterns 145
265